Scheduled Ookla speed tests, 8 Feb – 22 Nov 2026 · 24,909 server measurements over 282 days with data · times in Europe/Helsinki · all rates in decimal Mbps
This report documents a long-running problem on a LocalNet (Local Loop Oy) 1000/1000 Mbps fiber connection: for much of 2026, evening throughput toward most of the internet collapsed, with packet loss hurting VoIP and remote work. SuomiCom downloads held line rate: the bottleneck was upstream, not the access line.
Until 22 November, Ookla speed tests ran 9 times a day (7 before 20 August) against the same four Helsinki-area servers, from four sites behind identical routers:
The short version (details in the timeline):
| Slot | LocalNet down (Mbps) | Bahnhof down (Mbps) | Elisa down (Mbps) | DNA down (Mbps) | LocalNet down, 1–29 Oct (Mbps) | LocalNet up (Mbps) | Bahnhof up (Mbps) | Elisa up (Mbps) | DNA up (Mbps) |
|---|---|---|---|---|---|---|---|---|---|
| 01:15 | 928 | 929 | 928 | 583 | 926 | 938 | 926 | 779 | 98 |
| 07:15 | 928 | 930 | 940 | 585 | 926 | 938 | 926 | 788 | 98 |
| 13:15 | 926 | 930 | 933 | 577 | 924 | 938 | 926 | 782 | 96 |
| 17:15 | 925 | 929 | 930 | 564 | 904 | 937 | 926 | 782 | 96 |
| 19:15 | 916 | 929 | 929 | 545 | 296 | 933 | 926 | 785 | 94 |
| 20:15 | 704 | 928 | 927 | 520 | 120 | 371 | 926 | 761 | 91 |
| 21:15 | 715 | 930 | 925 | 519 | 84 | 904 | 926 | 785 | 91 |
| 22:15 | 558 | 928 | 924 | 530 | 149 | 344 | 926 | 762 | 89 |
| 23:15 | 923 | 928 | 923 | 551 | 902 | 936 | 926 | 781 | 92 |
All four ISPs deliver ~92–97% of plan off-peak. At 21:15 LocalNet's median falls to 71.5% — and that's the median over the whole period, including the good months. DNA dips mildly (to 86.6%); Elisa barely moves, and Bahnhof's runs since 14 September are flat at ~93%. At its worst, in the four weeks before the router update, LocalNet's median was 30% at 19:15 and 8% at 21:15, still 15% at 22:15, and back to 90% by 23:15.
| Server | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov |
|---|---|---|---|---|---|---|---|---|---|---|
| SuomiCom (Espoo) | 936 | 938 | 938 | 938 | 938 | 938 | 938 | 938 | 938 | 939 |
| DNA Welho (Helsinki) | 323 | 350 | 829 | 760 | 884 | 902 | 720 | 228 | 125 | 920 |
| Telia (Helsinki) | 356 | 494 | 883 | 867 | 902 | 918 | 737 | 257 | 136 | 925 |
| Elisa (Helsinki) | 390 | 510 | 878 | 898 | 911 | 918 | 851 | 332 | 123 | 922 |
The tell: tests toward SuomiCom (Espoo) ran at ~937 Mbps in every month — the access line itself was never the problem. Congestion was in LocalNet's interconnects toward the rest of the internet (DNA Welho, Telia and Elisa directions), matching the ISP's own diagnosis of saturated upstream/metro capacity.
| Date | Median Mbps | Min Mbps | Max Mbps |
|---|---|---|---|
| 2026-02-08 | 78 | 49 | 927 |
| 2026-02-09 | 257 | 33 | 935 |
| 2026-02-10 | 91 | 5 | 927 |
| 2026-02-11 | 891 | 68 | 940 |
| 2026-02-12 | 585 | 54 | 936 |
| 2026-02-13 | 816 | 353 | 939 |
| 2026-02-14 | 551 | 204 | 936 |
| 2026-02-15 | 522 | 166 | 940 |
| 2026-02-16 | 898 | 144 | 938 |
| 2026-02-17 | 421 | 97 | 940 |
| 2026-02-18 | 327 | 213 | 938 |
| 2026-02-19 | 326 | 72 | 932 |
| 2026-02-20 | 508 | 188 | 937 |
| 2026-02-21 | 611 | 144 | 936 |
| 2026-02-22 | 528 | 256 | 945 |
| 2026-02-23 | 598 | 387 | 936 |
| 2026-02-24 | 506 | 282 | 938 |
| 2026-02-25 | 578 | 280 | 941 |
| 2026-02-26 | 908 | 323 | 939 |
| 2026-02-27 | 808 | 79 | 937 |
| 2026-02-28 | 841 | 124 | 940 |
| 2026-03-01 | 399 | 149 | 941 |
| 2026-03-02 | 860 | 251 | 939 |
| 2026-03-03 | 884 | 225 | 941 |
| 2026-03-04 | 882 | 336 | 938 |
| 2026-03-05 | 901 | 152 | 940 |
| 2026-03-06 | 900 | 196 | 940 |
| 2026-03-07 | 539 | 278 | 938 |
| 2026-03-08 | 435 | 161 | 938 |
| 2026-03-09 | 913 | 241 | 931 |
| 2026-03-10 | 896 | 259 | 939 |
| 2026-03-11 | 802 | 259 | 938 |
| 2026-03-12 | 884 | 345 | 939 |
| 2026-03-13 | 535 | 273 | 938 |
| 2026-03-14 | 389 | 120 | 939 |
| 2026-03-15 | 258 | 107 | 939 |
| 2026-03-16 | 787 | 431 | 939 |
| 2026-03-17 | 904 | 604 | 937 |
| 2026-03-18 | 905 | 197 | 940 |
| 2026-03-19 | 837 | 185 | 939 |
| 2026-03-20 | 626 | 298 | 941 |
| 2026-03-21 | 200 | 112 | 940 |
| 2026-03-22 | 263 | 56 | 939 |
| 2026-03-23 | 289 | 156 | 940 |
| 2026-03-24 | 252 | 183 | 939 |
| 2026-03-25 | 736 | 282 | 939 |
| 2026-03-26 | 809 | 202 | 939 |
| 2026-03-27 | 818 | 302 | 940 |
| 2026-03-28 | 882 | 288 | 938 |
| 2026-03-29 | 818 | 283 | 934 |
| 2026-03-30 | 864 | 298 | 937 |
| 2026-03-31 | 877 | 171 | 941 |
| 2026-04-01 | 900 | 98 | 938 |
| 2026-04-02 | 902 | 298 | 937 |
| 2026-04-03 | 914 | 858 | 938 |
| 2026-04-04 | 912 | 796 | 940 |
| 2026-04-05 | 884 | 351 | 940 |
| 2026-04-06 | 596 | 118 | 940 |
| 2026-04-07 | 923 | 847 | 938 |
| 2026-04-08 | 877 | 127 | 939 |
| 2026-04-09 | 915 | 266 | 938 |
| 2026-04-10 | 888 | 403 | 940 |
| 2026-04-11 | 851 | 510 | 939 |
| 2026-04-12 | 900 | 495 | 938 |
| 2026-04-13 | 916 | 520 | 941 |
| 2026-04-14 | 911 | 239 | 940 |
| 2026-04-15 | 904 | 207 | 938 |
| 2026-04-16 | 694 | 158 | 938 |
| 2026-04-17 | 914 | 210 | 940 |
| 2026-04-18 | 879 | 772 | 942 |
| 2026-04-19 | 393 | 259 | 937 |
| 2026-04-20 | 919 | 366 | 940 |
| 2026-04-21 | 915 | 740 | 941 |
| 2026-04-22 | 917 | 884 | 939 |
| 2026-04-23 | 924 | 832 | 938 |
| 2026-04-24 | 924 | 834 | 938 |
| 2026-04-25 | 848 | 613 | 939 |
| 2026-04-26 | 428 | 278 | 939 |
| 2026-04-27 | 543 | 324 | 935 |
| 2026-04-28 | 914 | 225 | 938 |
| 2026-04-29 | 912 | 827 | 936 |
| 2026-04-30 | 915 | 889 | 938 |
| 2026-05-01 | 926 | 896 | 936 |
| 2026-05-02 | 918 | 714 | 940 |
| 2026-05-03 | 875 | 139 | 938 |
| 2026-05-04 | 923 | 300 | 940 |
| 2026-05-05 | 809 | 266 | 939 |
| 2026-05-06 | 914 | 330 | 938 |
| 2026-05-07 | 923 | 542 | 939 |
| 2026-05-08 | 910 | 373 | 939 |
| 2026-05-09 | 891 | 170 | 938 |
| 2026-05-10 | 818 | 126 | 940 |
| 2026-05-11 | 899 | 384 | 939 |
| 2026-05-12 | 552 | 133 | 939 |
| 2026-05-13 | 830 | 66 | 937 |
| 2026-05-14 | 922 | 463 | 938 |
| 2026-05-15 | 923 | 773 | 939 |
| 2026-05-16 | 920 | 669 | 939 |
| 2026-05-17 | 891 | 592 | 938 |
| 2026-05-18 | 921 | 674 | 943 |
| 2026-05-19 | 908 | 734 | 938 |
| 2026-05-20 | 910 | 661 | 939 |
| 2026-05-21 | 522 | 263 | 939 |
| 2026-05-22 | 835 | 292 | 938 |
| 2026-05-23 | 912 | 595 | 940 |
| 2026-05-24 | 915 | 377 | 938 |
| 2026-05-25 | 907 | 693 | 938 |
| 2026-05-26 | 905 | 599 | 939 |
| 2026-05-27 | 913 | 322 | 941 |
| 2026-05-28 | 757 | 167 | 938 |
| 2026-05-29 | 908 | 412 | 938 |
| 2026-05-30 | 920 | 717 | 942 |
| 2026-05-31 | 844 | 222 | 937 |
| 2026-06-01 | 913 | 330 | 937 |
| 2026-06-02 | 910 | 390 | 935 |
| 2026-06-03 | 926 | 722 | 940 |
| 2026-06-04 | 900 | 403 | 940 |
| 2026-06-05 | 920 | 241 | 944 |
| 2026-06-06 | 894 | 601 | 939 |
| 2026-06-07 | 910 | 788 | 942 |
| 2026-06-08 | 913 | 207 | 938 |
| 2026-06-09 | 516 | 189 | 939 |
| 2026-06-10 | 922 | 231 | 940 |
| 2026-06-11 | 888 | 182 | 937 |
| 2026-06-12 | 920 | 557 | 940 |
| 2026-06-13 | 914 | 243 | 937 |
| 2026-06-14 | 904 | 207 | 939 |
| 2026-06-15 | 783 | 617 | 939 |
| 2026-06-16 | 916 | 723 | 938 |
| 2026-06-17 | 922 | 804 | 938 |
| 2026-06-18 | 922 | 906 | 938 |
| 2026-06-19 | 927 | 911 | 935 |
| 2026-06-20 | 921 | 898 | 939 |
| 2026-06-21 | 916 | 578 | 936 |
| 2026-06-22 | 910 | 243 | 940 |
| 2026-06-23 | 904 | 222 | 939 |
| 2026-06-24 | 918 | 789 | 941 |
| 2026-06-25 | 905 | 231 | 938 |
| 2026-06-26 | 913 | 636 | 938 |
| 2026-06-27 | 921 | 893 | 938 |
| 2026-06-28 | 888 | 372 | 939 |
| 2026-06-29 | 926 | 589 | 939 |
| 2026-06-30 | 906 | 296 | 938 |
| 2026-07-01 | 919 | 908 | 946 |
| 2026-07-02 | 913 | 479 | 938 |
| 2026-07-03 | 921 | 883 | 943 |
| 2026-07-04 | 907 | 580 | 938 |
| 2026-07-05 | 915 | 400 | 940 |
| 2026-07-06 | 924 | 902 | 939 |
| 2026-07-07 | 845 | 495 | 940 |
| 2026-07-08 | 924 | 879 | 938 |
| 2026-07-09 | 920 | 835 | 938 |
| 2026-07-10 | 932 | 902 | 940 |
| 2026-07-11 | 927 | 898 | 939 |
| 2026-07-12 | 904 | 444 | 938 |
| 2026-07-13 | 922 | 898 | 938 |
| 2026-07-14 | 922 | 898 | 939 |
| 2026-07-15 | 932 | 912 | 939 |
| 2026-07-16 | 926 | 870 | 940 |
| 2026-07-17 | 917 | 882 | 938 |
| 2026-07-18 | 918 | 892 | 938 |
| 2026-07-20 | 911 | 894 | 942 |
| 2026-07-21 | 917 | 621 | 937 |
| 2026-07-22 | 921 | 860 | 939 |
| 2026-07-23 | 928 | 896 | 938 |
| 2026-07-24 | 914 | 838 | 935 |
| 2026-07-25 | 897 | 611 | 933 |
| 2026-07-26 | 626 | 295 | 938 |
| 2026-07-27 | 926 | 233 | 940 |
| 2026-07-28 | 904 | 494 | 938 |
| 2026-07-29 | 921 | 913 | 938 |
| 2026-07-30 | 925 | 905 | 938 |
| 2026-07-31 | 919 | 848 | 943 |
| 2026-08-01 | 930 | 904 | 941 |
| 2026-08-02 | 916 | 854 | 939 |
| 2026-08-03 | 904 | 530 | 938 |
| 2026-08-04 | 922 | 872 | 940 |
| 2026-08-05 | 910 | 345 | 943 |
| 2026-08-06 | 916 | 327 | 938 |
| 2026-08-07 | 909 | 414 | 937 |
| 2026-08-08 | 916 | 881 | 936 |
| 2026-08-09 | 922 | 138 | 937 |
| 2026-08-10 | 841 | 137 | 937 |
| 2026-08-11 | 605 | 118 | 943 |
| 2026-08-19 | 910 | 296 | 939 |
| 2026-08-20 | 390 | 120 | 940 |
| 2026-08-21 | 707 | 337 | 942 |
| 2026-08-22 | 843 | 63 | 940 |
| 2026-08-23 | 299 | 135 | 940 |
| 2026-08-24 | 833 | 183 | 938 |
| 2026-08-25 | 906 | 179 | 943 |
| 2026-08-26 | 811 | 159 | 938 |
| 2026-08-27 | 874 | 123 | 937 |
| 2026-08-28 | 835 | 252 | 940 |
| 2026-08-29 | 908 | 445 | 944 |
| 2026-08-30 | 220 | 73 | 938 |
| 2026-08-31 | 904 | 128 | 937 |
| 2026-09-01 | 854 | 283 | 940 |
| 2026-09-02 | 890 | 192 | 944 |
| 2026-09-03 | 903 | 117 | 938 |
| 2026-09-04 | 457 | 61 | 940 |
| 2026-09-05 | 707 | 248 | 938 |
| 2026-09-06 | 287 | 88 | 939 |
| 2026-09-07 | 826 | 437 | 940 |
| 2026-09-08 | 830 | 100 | 938 |
| 2026-09-09 | 636 | 67 | 935 |
| 2026-09-10 | 899 | 138 | 940 |
| 2026-09-11 | 800 | 81 | 944 |
| 2026-09-12 | 515 | 93 | 938 |
| 2026-09-14 | 876 | 81 | 938 |
| 2026-09-15 | 866 | 69 | 937 |
| 2026-09-16 | 170 | 65 | 942 |
| 2026-09-17 | 429 | 63 | 939 |
| 2026-09-18 | 413 | 86 | 939 |
| 2026-09-19 | 352 | 174 | 938 |
| 2026-09-20 | 243 | 106 | 940 |
| 2026-09-21 | 892 | 73 | 938 |
| 2026-09-22 | 755 | 87 | 938 |
| 2026-09-23 | 285 | 80 | 940 |
| 2026-09-24 | 532 | 69 | 939 |
| 2026-09-25 | 172 | 116 | 941 |
| 2026-09-26 | 398 | 128 | 939 |
| 2026-09-27 | 549 | 58 | 940 |
| 2026-09-28 | 813 | 62 | 940 |
| 2026-09-29 | 101 | 56 | 939 |
| 2026-09-30 | 481 | 98 | 941 |
| 2026-10-01 | 570 | 138 | 938 |
| 2026-10-02 | 255 | 66 | 939 |
| 2026-10-03 | 233 | 62 | 942 |
| 2026-10-04 | 325 | 59 | 940 |
| 2026-10-05 | 206 | 76 | 939 |
| 2026-10-06 | 560 | 74 | 940 |
| 2026-10-07 | 367 | 125 | 941 |
| 2026-10-08 | 162 | 75 | 935 |
| 2026-10-09 | 133 | 53 | 937 |
| 2026-10-10 | 284 | 135 | 939 |
| 2026-10-11 | 95 | 63 | 939 |
| 2026-10-12 | 725 | 69 | 939 |
| 2026-10-13 | 218 | 112 | 938 |
| 2026-10-14 | 709 | 62 | 940 |
| 2026-10-15 | 225 | 66 | 940 |
| 2026-10-16 | 149 | 63 | 943 |
| 2026-10-17 | 158 | 64 | 947 |
| 2026-10-18 | 94 | 61 | 940 |
| 2026-10-19 | 142 | 60 | 940 |
| 2026-10-20 | 162 | 50 | 937 |
| 2026-10-21 | 373 | 63 | 940 |
| 2026-10-22 | 371 | 62 | 940 |
| 2026-10-23 | 264 | 60 | 940 |
| 2026-10-24 | 171 | 75 | 939 |
| 2026-10-25 | 72 | 24 | 939 |
| 2026-10-26 | 281 | 62 | 938 |
| 2026-10-27 | 83 | 55 | 940 |
| 2026-10-28 | 96 | 62 | 939 |
| 2026-10-29 | 197 | 53 | 939 |
| 2026-10-30 | 920 | 897 | 940 |
| 2026-10-31 | 918 | 896 | 938 |
| 2026-11-01 | 918 | 902 | 940 |
| 2026-11-02 | 918 | 904 | 939 |
| 2026-11-03 | 926 | 911 | 940 |
| 2026-11-04 | 924 | 906 | 943 |
| 2026-11-05 | 918 | 899 | 941 |
| 2026-11-06 | 920 | 895 | 942 |
| 2026-11-07 | 923 | 907 | 942 |
| 2026-11-08 | 924 | 915 | 939 |
| 2026-11-09 | 923 | 890 | 941 |
| 2026-11-10 | 922 | 902 | 938 |
| 2026-11-11 | 918 | 902 | 939 |
| 2026-11-12 | 933 | 923 | 943 |
| 2026-11-13 | 930 | 920 | 941 |
| 2026-11-14 | 927 | 921 | 943 |
| 2026-11-15 | 930 | 915 | 940 |
| 2026-11-16 | 932 | 922 | 942 |
| 2026-11-17 | 930 | 890 | 938 |
| 2026-11-18 | 927 | 893 | 938 |
| 2026-11-19 | 931 | 908 | 940 |
| 2026-11-20 | 926 | 917 | 939 |
| 2026-11-21 | 928 | 906 | 940 |
February's median sat around 520 Mbps with the worst runs under 10 Mbps; recovery came stepwise after the mid-February fixes and the 23 March “Phase 1” migration, and in July the worst tenth of primetime runs stayed above 793 Mbps.
The regression since 9 August kept deepening through 29 October. From mid-October the collapse ran from 19:15 or 20:15 through 22:15 every night, on the worst from 17:15, and 23:15 often failed to recover; 25 and 27 October's daily primetime medians of 72 and 83 Mbps are the two lowest of any day with both slots tested, both under 10 February's 91, and 25 October's 21:15 median of 40 Mbps the lowest since that day. From 30 October, the day the router update completed, every evening ran at line rate in every slot until the tests ended, the first clean evenings since the new series began on 19 August. The midday collapses of 8 and 10 September and 15 October fall outside the two slots this chart plots.
| Date | Median Mbps | Min Mbps | Max Mbps |
|---|---|---|---|
| 2026-02-08 | 884 | 783 | 936 |
| 2026-02-09 | 898 | 821 | 941 |
| 2026-02-10 | 502 | 351 | 604 |
| 2026-02-11 | 911 | 909 | 925 |
| 2026-02-12 | 921 | 904 | 931 |
| 2026-02-13 | 933 | 930 | 937 |
| 2026-02-14 | 926 | 924 | 939 |
| 2026-02-15 | 926 | 920 | 935 |
| 2026-02-16 | 928 | 912 | 937 |
| 2026-02-17 | 925 | 911 | 932 |
| 2026-02-18 | 931 | 924 | 934 |
| 2026-02-19 | 667 | 414 | 935 |
| 2026-02-20 | 927 | 921 | 931 |
| 2026-02-21 | 927 | 914 | 934 |
| 2026-02-22 | 929 | 927 | 938 |
| 2026-02-23 | 929 | 924 | 935 |
| 2026-02-24 | 930 | 927 | 939 |
| 2026-02-25 | 927 | 925 | 936 |
| 2026-02-26 | 929 | 926 | 938 |
| 2026-02-27 | 912 | 858 | 931 |
| 2026-02-28 | 928 | 902 | 934 |
| 2026-03-01 | 930 | 925 | 937 |
| 2026-03-02 | 928 | 927 | 938 |
| 2026-03-03 | 930 | 921 | 934 |
| 2026-03-04 | 931 | 927 | 937 |
| 2026-03-05 | 930 | 928 | 938 |
| 2026-03-06 | 932 | 930 | 938 |
| 2026-03-07 | 929 | 924 | 938 |
| 2026-03-08 | 927 | 920 | 935 |
| 2026-03-09 | 926 | 916 | 934 |
| 2026-03-10 | 930 | 922 | 933 |
| 2026-03-11 | 927 | 924 | 931 |
| 2026-03-12 | 927 | 924 | 939 |
| 2026-03-13 | 928 | 925 | 930 |
| 2026-03-14 | 930 | 924 | 938 |
| 2026-03-15 | 928 | 925 | 936 |
| 2026-03-16 | 930 | 922 | 937 |
| 2026-03-17 | 926 | 910 | 941 |
| 2026-03-18 | 928 | 926 | 935 |
| 2026-03-19 | 930 | 923 | 936 |
| 2026-03-20 | 932 | 928 | 933 |
| 2026-03-21 | 923 | 913 | 929 |
| 2026-03-22 | 900 | 537 | 929 |
| 2026-03-23 | 930 | 926 | 932 |
| 2026-03-24 | 931 | 924 | 942 |
| 2026-03-25 | 934 | 931 | 934 |
| 2026-03-26 | 929 | 927 | 934 |
| 2026-03-27 | 932 | 927 | 938 |
| 2026-03-28 | 931 | 929 | 935 |
| 2026-03-29 | 928 | 925 | 939 |
| 2026-03-30 | 930 | 925 | 934 |
| 2026-03-31 | 929 | 920 | 935 |
| 2026-04-01 | 929 | 924 | 932 |
| 2026-04-02 | 930 | 927 | 941 |
| 2026-04-03 | 934 | 933 | 937 |
| 2026-04-04 | 931 | 924 | 932 |
| 2026-04-05 | 931 | 926 | 938 |
| 2026-04-06 | 931 | 923 | 934 |
| 2026-04-07 | 930 | 927 | 931 |
| 2026-04-08 | 910 | 837 | 932 |
| 2026-04-09 | 385 | 270 | 916 |
| 2026-04-10 | 448 | 304 | 937 |
| 2026-04-11 | 692 | 594 | 935 |
| 2026-04-12 | 440 | 370 | 938 |
| 2026-04-13 | 622 | 439 | 936 |
| 2026-04-14 | 317 | 177 | 937 |
| 2026-04-15 | 234 | 216 | 938 |
| 2026-04-16 | 274 | 190 | 938 |
| 2026-04-17 | 263 | 215 | 938 |
| 2026-04-18 | 873 | 725 | 937 |
| 2026-04-19 | 252 | 243 | 936 |
| 2026-04-20 | 472 | 379 | 937 |
| 2026-04-21 | 644 | 472 | 938 |
| 2026-04-22 | 902 | 877 | 938 |
| 2026-04-23 | 847 | 679 | 938 |
| 2026-04-24 | 849 | 657 | 937 |
| 2026-04-25 | 611 | 486 | 938 |
| 2026-04-26 | 290 | 242 | 933 |
| 2026-04-27 | 319 | 233 | 937 |
| 2026-04-28 | 269 | 203 | 935 |
| 2026-04-29 | 922 | 891 | 938 |
| 2026-04-30 | 937 | 889 | 946 |
| 2026-05-01 | 937 | 921 | 938 |
| 2026-05-02 | 742 | 486 | 938 |
| 2026-05-03 | 161 | 119 | 937 |
| 2026-05-04 | 442 | 317 | 938 |
| 2026-05-05 | 372 | 232 | 938 |
| 2026-05-06 | 448 | 306 | 938 |
| 2026-05-07 | 603 | 514 | 938 |
| 2026-05-08 | 453 | 341 | 938 |
| 2026-05-09 | 288 | 213 | 936 |
| 2026-05-10 | 230 | 193 | 935 |
| 2026-05-11 | 388 | 364 | 936 |
| 2026-05-12 | 251 | 122 | 938 |
| 2026-05-13 | 100 | 82 | 914 |
| 2026-05-14 | 378 | 287 | 937 |
| 2026-05-15 | 798 | 583 | 938 |
| 2026-05-16 | 846 | 583 | 938 |
| 2026-05-17 | 440 | 280 | 938 |
| 2026-05-18 | 803 | 487 | 934 |
| 2026-05-19 | 693 | 425 | 938 |
| 2026-05-20 | 695 | 522 | 929 |
| 2026-05-21 | 347 | 226 | 938 |
| 2026-05-22 | 476 | 419 | 928 |
| 2026-05-23 | 462 | 412 | 936 |
| 2026-05-24 | 453 | 399 | 937 |
| 2026-05-25 | 765 | 699 | 938 |
| 2026-05-26 | 490 | 397 | 937 |
| 2026-05-27 | 349 | 266 | 938 |
| 2026-05-28 | 275 | 126 | 930 |
| 2026-05-29 | 373 | 308 | 938 |
| 2026-05-30 | 531 | 336 | 938 |
| 2026-05-31 | 247 | 176 | 938 |
| 2026-06-01 | 308 | 270 | 938 |
| 2026-06-02 | 432 | 396 | 938 |
| 2026-06-03 | 746 | 399 | 938 |
| 2026-06-04 | 521 | 352 | 907 |
| 2026-06-05 | 291 | 209 | 938 |
| 2026-06-06 | 652 | 592 | 938 |
| 2026-06-07 | 678 | 536 | 938 |
| 2026-06-08 | 216 | 184 | 935 |
| 2026-06-09 | 430 | 323 | 938 |
| 2026-06-10 | 344 | 276 | 936 |
| 2026-06-11 | 204 | 142 | 936 |
| 2026-06-12 | 624 | 339 | 938 |
| 2026-06-13 | 380 | 303 | 938 |
| 2026-06-14 | 203 | 168 | 938 |
| 2026-06-15 | 392 | 285 | 938 |
| 2026-06-16 | 717 | 546 | 938 |
| 2026-06-17 | 494 | 413 | 936 |
| 2026-06-18 | 938 | 935 | 943 |
| 2026-06-19 | 938 | 935 | 945 |
| 2026-06-20 | 938 | 936 | 946 |
| 2026-06-21 | 484 | 400 | 933 |
| 2026-06-22 | 230 | 218 | 938 |
| 2026-06-23 | 234 | 183 | 928 |
| 2026-06-24 | 897 | 554 | 937 |
| 2026-06-25 | 337 | 285 | 937 |
| 2026-06-26 | 775 | 416 | 941 |
| 2026-06-27 | 937 | 936 | 944 |
| 2026-06-28 | 318 | 255 | 938 |
| 2026-06-29 | 625 | 496 | 936 |
| 2026-06-30 | 315 | 278 | 939 |
| 2026-07-01 | 937 | 930 | 940 |
| 2026-07-02 | 530 | 421 | 939 |
| 2026-07-03 | 806 | 562 | 938 |
| 2026-07-04 | 540 | 324 | 938 |
| 2026-07-05 | 302 | 270 | 937 |
| 2026-07-06 | 930 | 897 | 938 |
| 2026-07-07 | 480 | 336 | 939 |
| 2026-07-08 | 926 | 887 | 939 |
| 2026-07-09 | 935 | 906 | 945 |
| 2026-07-10 | 938 | 931 | 945 |
| 2026-07-11 | 938 | 935 | 946 |
| 2026-07-12 | 366 | 312 | 937 |
| 2026-07-13 | 868 | 719 | 938 |
| 2026-07-14 | 938 | 891 | 943 |
| 2026-07-15 | 938 | 926 | 946 |
| 2026-07-16 | 937 | 932 | 938 |
| 2026-07-17 | 935 | 796 | 945 |
| 2026-07-20 | 864 | 323 | 937 |
| 2026-07-21 | 420 | 332 | 939 |
| 2026-07-22 | 757 | 615 | 938 |
| 2026-07-23 | 936 | 916 | 939 |
| 2026-07-24 | 842 | 703 | 931 |
| 2026-07-25 | 658 | 542 | 937 |
| 2026-07-26 | 178 | 144 | 934 |
| 2026-07-27 | 445 | 222 | 938 |
| 2026-07-28 | 459 | 413 | 936 |
| 2026-07-29 | 936 | 829 | 942 |
| 2026-07-30 | 906 | 695 | 942 |
| 2026-07-31 | 766 | 657 | 938 |
| 2026-08-01 | 936 | 926 | 940 |
| 2026-08-02 | 696 | 520 | 936 |
| 2026-08-03 | 449 | 295 | 938 |
| 2026-08-04 | 802 | 692 | 938 |
| 2026-08-05 | 280 | 184 | 939 |
| 2026-08-06 | 310 | 276 | 938 |
| 2026-08-07 | 405 | 281 | 935 |
| 2026-08-08 | 707 | 669 | 937 |
| 2026-08-09 | 173 | 140 | 938 |
| 2026-08-10 | 146 | 111 | 938 |
| 2026-08-11 | 130 | 102 | 935 |
| 2026-08-19 | 339 | 313 | 937 |
| 2026-08-20 | 129 | 102 | 939 |
| 2026-08-21 | 318 | 296 | 938 |
| 2026-08-22 | 73 | 55 | 938 |
| 2026-08-23 | 151 | 117 | 937 |
| 2026-08-24 | 151 | 107 | 938 |
| 2026-08-25 | 180 | 146 | 904 |
| 2026-08-26 | 266 | 174 | 938 |
| 2026-08-27 | 220 | 162 | 936 |
| 2026-08-28 | 329 | 217 | 936 |
| 2026-08-29 | 391 | 333 | 934 |
| 2026-08-30 | 81 | 64 | 937 |
| 2026-08-31 | 160 | 117 | 938 |
| 2026-09-01 | 226 | 155 | 938 |
| 2026-09-02 | 324 | 216 | 934 |
| 2026-09-03 | 125 | 91 | 937 |
| 2026-09-04 | 62 | 61 | 937 |
| 2026-09-05 | 280 | 232 | 937 |
| 2026-09-06 | 129 | 116 | 938 |
| 2026-09-07 | 403 | 320 | 905 |
| 2026-09-08 | 108 | 107 | 938 |
| 2026-09-09 | 70 | 65 | 853 |
| 2026-09-10 | 201 | 156 | 937 |
| 2026-09-11 | 102 | 88 | 936 |
| 2026-09-12 | 111 | 78 | 939 |
| 2026-09-14 | 102 | 89 | 939 |
| 2026-09-15 | 79 | 75 | 927 |
| 2026-09-16 | 64 | 56 | 897 |
| 2026-09-17 | 67 | 64 | 934 |
| 2026-09-18 | 152 | 96 | 937 |
| 2026-09-19 | 177 | 156 | 939 |
| 2026-09-20 | 109 | 89 | 938 |
| 2026-09-21 | 90 | 68 | 938 |
| 2026-09-22 | 149 | 129 | 914 |
| 2026-09-23 | 134 | 114 | 938 |
| 2026-09-24 | 72 | 65 | 532 |
| 2026-09-25 | 177 | 150 | 933 |
| 2026-09-26 | 146 | 124 | 565 |
| 2026-09-27 | 74 | 55 | 931 |
| 2026-09-28 | 71 | 51 | 623 |
| 2026-09-29 | 100 | 78 | 286 |
| 2026-09-30 | 115 | 99 | 938 |
| 2026-10-01 | 139 | 129 | 933 |
| 2026-10-02 | 101 | 72 | 937 |
| 2026-10-03 | 75 | 69 | 938 |
| 2026-10-04 | 72 | 62 | 939 |
| 2026-10-05 | 89 | 74 | 888 |
| 2026-10-06 | 62 | 58 | 824 |
| 2026-10-07 | 131 | 120 | 939 |
| 2026-10-08 | 110 | 84 | 938 |
| 2026-10-09 | 77 | 61 | 905 |
| 2026-10-10 | 244 | 187 | 937 |
| 2026-10-11 | 79 | 62 | 937 |
| 2026-10-12 | 90 | 62 | 938 |
| 2026-10-13 | 126 | 92 | 936 |
| 2026-10-14 | 69 | 61 | 930 |
| 2026-10-15 | 63 | 60 | 938 |
| 2026-10-16 | 79 | 59 | 938 |
| 2026-10-17 | 89 | 56 | 938 |
| 2026-10-18 | 53 | 42 | 939 |
| 2026-10-19 | 65 | 62 | 663 |
| 2026-10-20 | 89 | 69 | 906 |
| 2026-10-21 | 63 | 61 | 939 |
| 2026-10-22 | 82 | 66 | 928 |
| 2026-10-23 | 100 | 94 | 878 |
| 2026-10-24 | 92 | 71 | 938 |
| 2026-10-25 | 50 | 39 | 774 |
| 2026-10-26 | 68 | 62 | 452 |
| 2026-10-27 | 91 | 80 | 754 |
| 2026-10-28 | 74 | 52 | 543 |
| 2026-10-29 | 54 | 35 | 939 |
| 2026-10-30 | 938 | 935 | 938 |
| 2026-10-31 | 938 | 938 | 939 |
| 2026-11-01 | 938 | 937 | 939 |
| 2026-11-02 | 938 | 937 | 939 |
| 2026-11-03 | 939 | 938 | 939 |
| 2026-11-04 | 938 | 938 | 939 |
| 2026-11-05 | 939 | 938 | 939 |
| 2026-11-06 | 938 | 937 | 939 |
| 2026-11-07 | 938 | 937 | 938 |
| 2026-11-08 | 937 | 936 | 938 |
| 2026-11-09 | 939 | 934 | 939 |
| 2026-11-10 | 937 | 936 | 938 |
| 2026-11-11 | 938 | 936 | 938 |
| 2026-11-12 | 938 | 937 | 939 |
| 2026-11-13 | 938 | 938 | 938 |
| 2026-11-14 | 938 | 936 | 939 |
| 2026-11-15 | 938 | 938 | 939 |
| 2026-11-16 | 938 | 937 | 939 |
| 2026-11-17 | 939 | 937 | 939 |
| 2026-11-18 | 938 | 938 | 938 |
| 2026-11-19 | 938 | 936 | 939 |
| 2026-11-20 | 937 | 935 | 938 |
| 2026-11-21 | 938 | 936 | 939 |
Upload was clean while downloads were at their worst — then collapsed from 9 April, exactly as download recovered: the congestion moved rather than disappeared. That spring episode touched only the 21:15 slot; by 19:15 and 23:15 upload was back at ~938 Mbps.
The August regression hits upload across the whole evening, and on the worst nights the 21:15 daily median falls to the 60s or below: 49.5 Mbps on 25 October, the lowest of the dataset, then 53.2 on 18 October and 53.5 on 29 October.
| Month | LocalNet median | Bahnhof median | Elisa median | DNA median | LocalNet p10 | Bahnhof p10 | Elisa p10 | DNA p10 |
|---|---|---|---|---|---|---|---|---|
| Feb | 318 | – | 914 | 510 | 67 | – | 886 | 414 |
| Mar | 383 | – | 921 | 511 | 170 | – | 901 | 403 |
| Apr | 825 | – | 918 | 521 | 245 | – | 900 | 304 |
| May | 714 | – | 919 | 510 | 195 | – | 896 | 373 |
| Jun | 825 | – | 928 | 527 | 252 | – | 903 | 422 |
| Jul | 914 | – | 930 | 534 | 566 | – | 918 | 442 |
| Aug | 537 | – | 928 | 519 | 128 | – | 908 | 392 |
| Sep | 141 | 929 | 927 | 512 | 68 | 921 | 914 | 400 |
| Oct | 87 | 930 | 929 | – | 61 | 920 | 912 | – |
| Nov | 926 | 930 | 920 | – | 904 | 920 | 905 | – |
LocalNet climbs from 32% of plan in February to 91% in July, while Elisa (91–93%) and DNA (85–89%) barely change. The improvement is real, but even July doesn't reach the controls' consistency. August's regression drags the 21:15 median back to 54%, with the worst tenth below 128 Mbps. September sits lower still: a 21:15 median of 141 Mbps, 14% of plan, below even February, and October lower again: 87 Mbps, 9% of plan. November's evenings, after the router update, read 93%. Bahnhof's evenings since 14 September read 93%.
| Week | Tests ≥1% loss | Mean loss | Max loss |
|---|---|---|---|
| 2026-02-09 | 14.6% | 0.67% | 14.2% |
| 2026-02-16 | 5.4% | 0.15% | 1.5% |
| 2026-02-23 | 3.4% | 0.09% | 1.3% |
| 2026-03-02 | 2.0% | 0.07% | 1.3% |
| 2026-03-09 | 2.7% | 0.1% | 1.6% |
| 2026-03-16 | 4.1% | 0.09% | 1.9% |
| 2026-03-23 | 0.7% | 0.07% | 1.3% |
| 2026-03-30 | 8.8% | 0.19% | 3.7% |
| 2026-04-06 | 7.5% | 0.17% | 2.7% |
| 2026-04-13 | 5.0% | 0.1% | 2.8% |
| 2026-04-20 | 4.8% | 0.1% | 2.6% |
| 2026-04-27 | 6.8% | 0.13% | 2.2% |
| 2026-05-04 | 4.8% | 0.1% | 2.0% |
| 2026-05-11 | 4.8% | 0.1% | 2.0% |
| 2026-05-18 | 8.2% | 0.17% | 3.8% |
| 2026-05-25 | 10.9% | 0.21% | 3.4% |
| 2026-06-01 | 7.5% | 0.15% | 2.5% |
| 2026-06-08 | 6.2% | 0.11% | 2.2% |
| 2026-06-15 | 3.4% | 0.08% | 2.5% |
| 2026-06-22 | 0.0% | 0.03% | 0.6% |
| 2026-06-29 | 4.1% | 0.1% | 1.7% |
| 2026-07-06 | 2.7% | 0.05% | 1.8% |
| 2026-07-13 | 1.7% | 0.03% | 1.3% |
| 2026-07-20 | 2.1% | 0.06% | 1.5% |
| 2026-07-27 | 0.7% | 0.04% | 1.3% |
| 2026-08-03 | 4.1% | 0.09% | 1.5% |
| 2026-08-10 | 2.3% | 0.07% | 1.0% |
| 2026-08-17 | 4.2% | 0.12% | 1.61% |
| 2026-08-24 | 2.1% | 0.07% | 1.22% |
| 2026-08-31 | 2.6% | 0.1% | 1.75% |
| 2026-09-07 | 7.7% | 0.22% | 9.13% |
| 2026-09-14 | 6.7% | 0.19% | 1.8% |
| 2026-09-21 | 5.9% | 0.36% | 23.33% |
| 2026-09-28 | 6.0% | 0.16% | 1.79% |
| 2026-10-05 | 3.2% | 0.16% | 2.38% |
| 2026-10-12 | 8.6% | 0.22% | 1.9% |
| 2026-10-19 | 6.9% | 0.23% | 1.94% |
| 2026-10-26 | 1.6% | 0.11% | 2.06% |
| 2026-11-02 | 0.0% | 0.0% | 0.43% |
| 2026-11-09 | 0.0% | 0% | 0% |
| 2026-11-16 | 0.0% | 0.01% | 0.93% |
Loss follows the same arc as throughput: worst in February (peaking at 14.6% of tests, single runs up to 14.2% loss), tapering through spring, near-zero after late June, and up again since August — the week of 12 October, at 8.6% of tests, was the highest share since late May.
In the evenings the regression mostly shows up as lost throughput rather than lost packets: evening runs of the new series stayed at or below 1.8% loss while speeds collapsed, until the 20:15 upload failure on 26 September read 23.3%, above February's worst; only five evening runs have passed 1.8% since: 2.4% toward SuomiCom at 17:15 on 7 October, as that direction's upload fell to 312 Mbps, and four of up to 2.1%, all in collapsed slots. The midday events of 8 and 10 September also carried loss — 9.1% toward SuomiCom on 8 September while that direction's upload collapsed, and 2.3% on the 10th. Across the whole period 4.51% of LocalNet tests saw ≥1% loss; Elisa and DNA logged one such test each (0.02%), Bahnhof none.
| Week | LocalNet-related (slots) | LocalNet-related (raised) | All other (slots) |
|---|---|---|---|
| 2026-03-09 | 31 | 29 | 2 |
| 2026-03-16 | 568 | 441 | 9 |
| 2026-03-23 | 113 | 107 | 13 |
| 2026-03-30 | 328 | 317 | 30 |
| 2026-04-06 | 283 | 245 | 9 |
| 2026-04-13 | 255 | 239 | 9 |
| 2026-04-20 | 259 | 239 | 701 |
| 2026-04-27 | 271 | 258 | 42 |
| 2026-05-04 | 314 | 306 | 11 |
| 2026-05-11 | 351 | 318 | 208 |
| 2026-05-18 | 451 | 420 | 581 |
| 2026-05-25 | 335 | 314 | 17 |
| 2026-06-01 | 299 | 289 | 8 |
| 2026-06-08 | 426 | 409 | 20 |
| 2026-06-15 | 232 | 209 | 0 |
| 2026-06-22 | 60 | 57 | 1,484 |
| 2026-06-29 | 92 | 89 | 0 |
| 2026-07-06 | 48 | 46 | 6 |
| 2026-07-13 | 69 | 69 | 18 |
| 2026-07-20 | 60 | 60 | 9 |
| 2026-07-27 | 83 | 78 | 0 |
| 2026-08-03 | 98 | 93 | 42 |
| 2026-08-10 | 221 | 195 | 36 |
| 2026-08-17 | 220 | 203 | 94 |
| 2026-08-24 | 222 | 197 | 32 |
| 2026-08-31 | 262 | 234 | 41 |
| 2026-09-07 | 407 | 342 | 37 |
| 2026-09-14 | 566 | 490 | 21 |
| 2026-09-21 | 536 | 475 | 15 |
| 2026-09-28 | 540 | 470 | 12 |
| 2026-10-05 | 710 | 597 | 9 |
| 2026-10-12 | 738 | 615 | 10 |
| 2026-10-19 | 932 | 775 | 804 |
| 2026-10-26 | 740 | 521 | 64 |
| 2026-11-02 | 104 | 38 | 8 |
| 2026-11-09 | 30 | 34 | 16 |
| 2026-11-16 | 27 | 13 | 62 |
A second, independent witness, measured as time in alert rather than alert counts, so sustained loss weighs what it lasted. LocalNet-related paths spent ~45 slots a day in alert until 18 June — peaking at 198 slots (16.5 hours) on 19 March — then fell off a cliff on 19 June: the same unannounced fix the speed tests see, dated to the day.
After the 9 August regression the time in alert tripled from July's ~10 to ~30 slots a day, bridging the 13–18 August speed-test gap and the 13 September client failure: the regression never paused. It kept climbing: the week of 19 October spent 932 slots in alert, the most of any week in the journal, and 29 October's 311 slots, 265 of them between 19:00 and midnight, make it the worst day of the whole journal by half again, past 25 October's 199, which had itself just edged out the March peak. The four days after it, 30 October – 2 November, are the journal's first with no LocalNet-related alert at all. The 94 slots on 3 November are not congestion: Telia's access-network maintenance in the Kauhava area that morning broke the line on and off for about an hour, and set off an instability in the monitoring network's own routing (see Method).
62% of the alert time falls between 19:00 and midnight. And 4,655 of the 9,831 alert episodes involve LocalNet's data center seen from vantage points outside the LocalNet line — the degradation is visible from the rest of the internet, not just from the customer's side.
| ISP | Idle (ms) | Loaded, download (ms) | Loaded, upload (ms) |
|---|---|---|---|
| LocalNet | 8.8 | 9.9 | 9.6 |
| Bahnhof | 16.1 | 18.4 | 17.9 |
| Elisa | 13.6 | 128.1 | 14.3 |
| DNA | 10.8 | 346.8 | 30.6 |
Outside the congested hours LocalNet's connection is excellent: latency stays at ~10 ms even under full download load. Elisa's rises to ~128 ms and DNA's to ~347 ms under load (bufferbloat), which hurts interactive use during large transfers.
Bahnhof, counted from its first runs behind the EdgeRouter 4 on 14 September, adds only ~2.3 ms under load, where LocalNet adds ~1.1 ms — the only other line here without bufferbloat. Its higher absolute latency is path length, not network quality: until 24 September Bahnhof idled about 1.5 ms above LocalNet, which is what the roughly 100 km longer fibre path from Ylivieska to Helsinki costs at the speed of light in glass, and since that day about 5 ms more, consistent with a longer route on Bahnhof's side (see the timeline).
| LocalNet | Bahnhof † | Elisa | DNA ‡ | |
|---|---|---|---|---|
| Subscribed (down/up, Mbps) | 1000 / 1000 | 1000 / 1000 | 1000 / 1000 | 600 / 100 |
| Median download, all slots | 923 Mbps | 929 Mbps | 930 Mbps | 566 Mbps |
| Median upload, all slots | 937 Mbps | 926 Mbps | 780 Mbps | 96 Mbps |
| Median download, primetime | 900 Mbps | 930 Mbps | 927 Mbps | 531 Mbps |
| 10th pct download, primetime | 142 Mbps | 920 Mbps | 906 Mbps | 423 Mbps |
| Tests with ≥1% packet loss | 4.51% | 0.0% | 0.02% | 0.02% |
| Median idle latency | 8.8 ms | 16.1 ms | 13.6 ms | 10.8 ms |
| Median loaded latency (down) | 9.9 ms | 18.4 ms | 128.1 ms | 346.8 ms |
| Measurements (server runs) | 8,436 | 2,454 | 8,354 | 5,661 |
From LocalNet's public status page (status.localnet.fi), the support thread with LocalNet (“Pakettikatoa iltaisin”, Jun 2025 – Jun 2026) and the measurement data. Status-page quotations are translated from Finnish.
From 7 February to 22 November 2026, a cron job ran the Ookla Speedtest CLI seven times daily — at 01:15, 07:15, 13:15, 17:15, 19:15, 21:15 and 23:15 (Europe/Helsinki); from 20 August two extra evening slots at 20:15 and 22:15 made it nine. Each run tested first the LocalNet connection, then Elisa (+1 min), then DNA (+2 min; until 7 September), and from 19 August also Bahnhof (+3 min). The scheduled runs were switched off on 22 November, once the line had been clean for 23 evenings; the last run is that day's 01:15.
The connections are at different sites — LocalNet in Kauhava (KAU), Elisa in Ylivieska (PRY), DNA in Espoo (ESP; decommissioned on 7 September), Bahnhof at a second Ylivieska site (TPY) — behind identical customer equipment (Ubiquiti EdgeRouter 4, gigabit LAN) — Bahnhof since the 12 September router swap, with its first runs behind that router on 14 September (see the timeline). All tested toward the same four fixed servers: SuomiCom (Espoo), DNA Welho, Telia and Elisa (all Helsinki). Differences in evening behavior therefore reflect the ISPs' networks, not the measurement setup; absolute idle latencies do differ with distance to Helsinki. LocalNet and Bahnhof are both delivered over the same open-access fibre network (Valokuitunen, built and operated by Telia, which the status page names as the party behind the access network, its handover links and their maintenance), which makes them the closest pair to compare and is why they sit side by side throughout.
ISP-side events are taken from LocalNet's public status page (status.localnet.fi, an Atlassian Statuspage whose incident and maintenance notices reach back to August 2024 and are mirrored in the repository) wherever it recorded them, and otherwise from the support thread, which also holds what the status page never published, such as the June fix.
The data comes in two series: the original text logs (8 February – 12 August, one file per run) and, after a 6-day pause, a new JSON series (from 19 August, one Ookla JSON result per server via speedtest -f json -b -s <id>). This report covers 8 February – 22 November 2026: 24,909 measurements; 33 scheduled runs in the old series produced no result (measurement-host gaps, e.g. 18–20 July) and are excluded, and the Ookla client failed over the weekend of 12–13 September: the 12 September evening runs are missing some servers, and no run completed between 13 September 07:15 and 14 September 17:15; on 29 September it failed again: no 19:15 runs, and at 20:15 only two LocalNet servers returned results. On 18 October the 07:15 LocalNet and Elisa runs produced no result, and on 21 October the Telia server returned no 21:15 result for any connection. Bahnhof's 856 runs behind its old router (19 August – 12 September) are excluded from every figure; its series starts 14 September 19:15. The connection under test is identified from the ISP field reported by the test itself.
Subscribed rates: LocalNet 1000/1000, Elisa 1000/1000, DNA 600/100, Bahnhof 1000/1000 Mbps. Units: the old text logs report rates in mebibits per second (“Mibps”), while the JSON series reports bytes/s (decimal). All old-series rates have been converted ×1.048576, so every figure in this report is in decimal Mbps — the unit Ookla and the ISPs use. (The initial 19 August publication mistakenly read the old logs as decimal Mbps, understating all rates by ~4.9%; relative comparisons were unaffected.)
Packet loss isn't reported by every server/run (6,248 of 24,909 measurements lack it — most often toward the DNA Welho server); loss statistics use only tests that reported it. Latency figures are Ookla's idle and loaded (during transfer) measurements; for the JSON series the loaded figure is the IQM. All aggregates are medians unless noted, computed per measurement (server run).
The alert chart comes from a separate SmokePing mesh, independent of the speed tests; its master sits on a different line entirely — a second, separate Elisa connection at the Ylivieska PR site. Every 5 minutes it pings its targets from seven vantage points: Kauhava (behind the LocalNet line), Lappeenranta (on a DNA connection), Avaruus.net and Hetzner in Helsinki, AWS Stockholm, Linode London, and the master's own site. Targets include each site's router, the routers' uplink gateways, hosts in LocalNet's data center, and DNS infrastructure in Finland and abroad.
An edge-triggered Loss alert is raised when a target shows non-zero packet loss in two polling rounds within 30 minutes, and stays raised while the pattern keeps matching (pattern >0%,*4*,>0%, unchanged since June 2025; the target set is unchanged since 11 March 2026, apart from the dead data-center host described below, removed on 23 August, the Espoo site's router, removed after that site was decommissioned on 7 September, and two data-center hosts that went silent late on 30 October, removed on the 31st; the April host and one of the October pair were put back at a restart at 23:56 that night).
The chart measures time in alert. Raised/cleared events are paired into episodes per target and vantage; each day's value is the number of 5-minute probe slots those episodes cover, summed over the paths that involve the LocalNet connection — measured from Kauhava, targeting the Kauhava routers or their uplinks, or targeting LocalNet's data center from any vantage point. Alerts still open when the SmokePing service restarts never log a “cleared”, so episodes are closed at service boundaries. A journal export is taken whenever the exporting script runs, which leaves the journal's first and last calendar days partial unless the export falls in a later month; the series covers complete local days only (15 Mar – 20 Nov here), so no bar stands for a fraction of a day.
Three exclusions keep the series about congestion. Dead hosts: one data-center test host went silent on 5 April and was next seen answering when it was put back into the target set on 31 October, the Espoo site's router went silent on 7 September when that site was decommissioned, and two more data-center hosts went silent together at about 23:45 on 30 October and were removed from the target set the next afternoon. A dead host sits in alert from every vantage until its target is removed, and re-raises after each service restart — an artifact, not congestion — so the April host is dropped, the Espoo router from 7 September on, and the two October hosts from the moment they went silent; the April host and the one October host put back at the 23:56 restart on 31 October are dropped only until then. And an episode is attributed to the far end when its target is simultaneously in alert from two or more non-Kauhava vantages — a 10-hour nameserver outage elsewhere on 26 June, or a 5-hour one on 25 October, is the far end's problem even when the Kauhava path also alerts on it. Such episodes count under “all other paths”. And the monitoring network's own routing: after a short line break during Telia maintenance early on 12 November, its internal routing over VPN tunnels went unstable, and the Kauhava probe's loss toward three of its targets until 10:52 that morning is dropped (194 slots). The same loss, toward the same three targets and from the Kauhava probe only, began inside Telia's maintenance window early on 18 November and lasted until 19:21 that evening; it is dropped likewise (527 slots), and so is every alert raised while all the monitoring sites' routers were rebooted around midnight on 20–21 November, for the firewall change meant to stop the instability. The issue also followed the 3 November break, but the line itself was going on and off that morning, so 3 November stays in the series.
Raised-alert counts remain in the table view for reference. The source is the SmokePing master's systemd journal, whose retention horizon is 14 March 2026 — the worse February period predates it and survives only in the complaint thread's counts.
Fixed Mbps vs Mibps confusion: the old logs are in mebibits/s, so all rates are now decimal Mbps (~4.9% higher than first published; comparisons unaffected).
New JSON test series (from 19 August) incorporated; Bahnhof added as a fourth connection.
The two new evening slots (20:15, 22:15) added to the slot table; the hourly chart keeps the original seven slots for comparability.
Added the SmokePing alert dataset (the monitoring master's journal, from 14 March) and the alert chart, measured as time in alert in 5-minute slots; a dead test host and far-end outages excluded, raise counts kept in the table view.
Sections restructured for readability: short figcaptions describe each figure, the analysis sits in body prose.
Next day: the alert series trimmed to complete local days; the “Loss alerts collapse” marker moved to 19 June, the day the drop happens.
Bahnhof's figures now count only runs from 14 September 19:15, behind its new router; its 856 router-capped runs are excluded from every figure.
The weekly loss chart omits weeks with fewer than a full test day of loss readings.
Prose condensed: the lead fits beside the KPI tiles, chart sections keep only what the figure cannot show, the timeline carries the history and the Bahnhof section is folded into it; changelog entries split into one paragraph per change, quiet refreshes merged.
Latency analysis: the Bahnhof–LocalNet gap attributed to the longer fibre path from Ylivieska (about 1.5 ms at idle, matching the distance) plus a small amount of queuing under load, replacing wording that read as a ranking of the two networks.
Bahnhof joins the hourly-profile and monthly 21:15 charts and their tables after a week of evenings behind its new router.
The connections are now ordered LocalNet, Bahnhof, Elisa, DNA throughout (the reason is in Method).
The “Slowest LocalNet run” tile, which only ever measured downloads, now shows the slowest upload beside the slowest download.
Latency analysis: Bahnhof's step in idle latency on 24 September attributed to a longer route; its load figure is now given as the delay added over idle.
Changelog trimmed to changes in the report itself; data refreshes no longer get entries.
Alert series: two data-center test hosts that went silent late on 30 October (and were removed from the target set the next day) are excluded from the moment they went silent, as the Espoo router was; without the exclusion their last hours would count as 19 slots of LocalNet-related alert on 30 October.
The alert chart's ceiling raised to hold 29 October.
LocalNet's public status page added as a source. Its incident and maintenance notices are mirrored into the repository, and the timeline now cites them where they cover an event, among them the December 2024 congestion, the August and October 2025 incidents, the February 2026 upgrade and the January–April maintenance nights, and the 28–30 October backbone router update, which also gets a chart marker; the “ISP fixes done” marker moves from 14 to 16 February, the day the status page closed that incident.
Alert series: when the journal is exported in the month after its last record, that record's day now counts as complete instead of being dropped as partial; 31 October joins the series.
The dead-host exclusions now end where a target was put back into the SmokePing target set (two of them at the 23:56 restart on 31 October, with no journal data after it yet), so a host that answers again counts from then on.
The hourly chart gains a dotted line: LocalNet's median over 1–29 October, the four weeks before its router update, also added to the table view. It runs through all nine slots, so the chart's axis now shows 20:15 and 22:15; the full-period lines still use the original seven.
The speed tests ended on 22 November; the lead, the timeline and Method now describe a closed series.
Alert series: the exclusion for the monitoring network's own routing instability now also covers 18 November, when the Kauhava probe's loss toward the same three targets ran from 03:11 to 19:21 (527 slots, which would otherwise be the journal's worst day), and the reboot of the monitoring sites' routers around midnight on 20–21 November.