Rolls25.9s
Scanner42.2s
FIGHT21.2s
Coverage3.8s
Incomenot reached
fleet peak285 msg/s
TWS CPU peak135.7 %

Latest open, per tab

tabgraded ontimeall rowsurgent rowswindow misses (first 90 s)lines peaksent 1st min (peak/s)err 101 / error 100 / reconnects
Rollseverything done25.9s25.9s13.9s0.0 %5043065 (225)0 / 0 / 0
Scannereverything done42.2s42.2s5.7s1.3 %2632638 (150)0 / 0 / 0
FIGHTeverything done21.2s21.2s21.2s-18226 (20)0 / 0 / 0
Coverageeverything done3.8s3.8s3.8s-013 (8)0 / 0 / 0
Incomeeverything donenot reached---0None (None)0 / 0 / 0

Every open (latest first)

ET dateGoal BRollsScannerFIGHTCoverageIncomeRolls / Scanner missesfleet peak msg/sTWS CPUwhat changed
2026-11-03not met25.9s42.2s21.2s3.8snot reached0.0 % / 1.3 %285135.7 %
2026-10-30PASS1m 47s1m 38s1m 47s8.9s8.5s0.6 % / 3.0 %331201.6 %
2026-10-29PASS1m 46s1m 37s1m 02s8.3s7.8s1.8 % / 3.2 %331191.7 %
2026-10-28PASS1m 52s1m 51s1m 00s10.4s8.2s0.9 % / 3.2 %368199.0 %
2026-10-27not met2m 46s2m 26s1m 24s12.7s10.6s0.8 % / 5.3 %344275.1 %Goal A PASS, Goal B not met: Rolls 2m 46s, Scanner 2m 26s; FIGHT 1m 24s (best yet, non-blocking verifies), Coverage 13s, Income 11s; 0 errors. Scanner picks 13s (picks fix worked). Cause: the contract-id cache was re-stamped ET 10-27 by off-hours saves, so yesterday's ~11k ids did not carry over and every tab re-qualified at the bell (Rolls 3,977 qualify messages in its first minute vs 0). Fixed for 10-28: the cache file keeps a carry section.
2026-10-26PASS1m 51s1m 53s1m 50s8.8s7.9s5.4 % / 3.7 %318167.2 %Goal B v2 PASS for the first time: Rolls 1m 51s, Scanner 1m 53s, FIGHT 1m 50s (11 fortresses), Coverage 9s, Income 8s; 0 errors, no wedge. Rolls no longer stalls (blocking re-plans wait for the whole catch-up). Scanner picks slower (1m 35s): rows left to Rolls by the one-fetcher wait.
2026-10-23not met2m 30s2m 32s2m 34s7.3s7.8s1.1 % / 0.6 %339337.1 %Goal A PASS, Goal B not met (Rolls 2m 30s, Scanner 2m 32s, FIGHT 2m 34s; Coverage 7s, Income 8s; 0 errors). Wins: FIGHT 2m 34s with 20 fortresses (3m 47s with 11 on 10-22) on its new bell windows; Scanner picks 5.2s (2m 00s on 10-22, bell re-ask); back-off 0 cuts (bell grace). Miss: Rolls stalled ~35 s from +65 s: three blocking CC-suggest re-plans (6.2 + 13.5 + 13.4 s) released at tier 2.
2026-10-22not met2m 31s2m 41s3m 47s11.5s15.9s5.5 % / 7.8 %335178.0 %Goal B changed to v2 tonight (owner): every tab everything done within 2 min. Goal A PASS. Real open ran 30-40 s slower than the rehearsals: 2.5x the tick traffic, every row moved overnight; the back-off cut Rolls at +5 s (options still opening), the shown-only hold ran out at +90 s and ~2,200 re-checks duplicated the catch-up, Scanner sat on 40 msg/s, FIGHT ran its leftover pieces one per cycle. Five fixes built for 10-23 (bell grace, Rolls hold-all, bell hand-off, bell re-ask, FIGHT burst to done).
2026-10-21not metnot reachednot reachednot reached10m 42snot reached- / -11391.2 %Not a test of the new levers: TWS was jammed before the bell. The Sheet export changed active_fortresses.csv (BMNR / MARA rows), so Rolls rejected its saved state at 21:20 and cold-downloaded 10,982 contracts at full speed (54,000 messages in 7 min). TWS memory went 2.0 -> 3.0 GB of its 4 GB limit, CPU 1,394%, and it stopped answering at ~21:23; restarted 21:40, watchers back from 21:45.
2026-10-20not met12m 32s3m 05s2m 09s9.9s30.5s91.4 % / 80.7 %591227.6 %Rolls fixes on (bell re-plan hold, non-blocking roll chain), FIGHT burst + Rolls quote reuse, Scanner at 250/s with a guaranteed 250-line pool. Lines were fine (0 err 101) but quotes stopped coming back: 91% of Rolls' and 81% of Scanner's window requests got no quote (10-19: 32% / 23%) while the fleet sent up to 591 msg/s. Next: Scanner back to 150/s at the bell (owner, 22:05 SGT).
2026-10-19not met5m 00s2m 39s2m 00s11.2s16.6s32.1 % / 23.1 %451179.1 %Scanner line pool switched on at the bell: Scanner 6m 30s down to 2m 39s. Rolls regressed to 5 min: its loop blocked in roll-chain fetches while holding ~790 lines, which also starved Scanner's pool.
2026-10-16not met1m 19s6m 30s54.1s15.9s22.5s7.3 % / 3.5 %434281.5 %First open with every watcher started before the bell. Rolls showed its rows in 79 s; Scanner took 6m 30s because its sweep ran one small window at a time.

Rehearsals (fake bell, mid-session)

Rolls, Scanner and FIGHT restarted with a pretend bell ~90 s out, graded by the same grader. Calmer than a real open, so times are slightly optimistic.

dateroundfake bellchangeRollsScannerFIGHTall <= 60 smisses R / Sfleet peakTWS CPUerrors
2026-10-22R001:17baseline, lever 1 off93 s122 s59 sno0 % / 0 %306153.40/0/0
2026-10-22R101:24lever 1 on95 s128 s74 sno0 % / 0 %304168.70/0/0
2026-10-22R201:30lever 1 + first/move verifies wait92 s123 s68 sno0 % / 0 %318220.70/0/0
2026-10-22R301:37R2 + more budget (WATCH_BELL_RATE_FIGHT=20, WATCH_BELL_RATE_REBUILD=250, WATCH_BELL_RATE_SCANNER=50)91 s85 s86 sno0 % / 0 %341165.30/0/0
2026-10-22R401:46R3 + Rolls catch-up pump (WATCH_BELL_RATE_FIGHT=25, WATCH_BELL_RATE_REBUILD=250, WATCH_BELL_RATE_SCANNER=50)-207 s109 sno63 % / 60 %334173.00/0/0
2026-10-22R501:52pump + guard fix, budget 285 (WATCH_BELL_RATE_FIGHT=25, WATCH_BELL_RATE_REBUILD=220, WATCH_BELL_RATE_SCANNER=40)91 s112 s75 sno0 % / 0 %333173.00/0/0
2026-10-22R601:59pump, Rolls 240 (WATCH_BELL_RATE_FIGHT=25, WATCH_BELL_RATE_REBUILD=240, WATCH_BELL_RATE_SCANNER=35)86 s126 s74 sno0 % / 0 %336178.80/0/0
2026-10-22R702:06pump, 230/50/20, no-pick fix (WATCH_BELL_RATE_FIGHT=20, WATCH_BELL_RATE_REBUILD=230, WATCH_BELL_RATE_SCANNER=50)91 s100 s84 sno6 % / 4 %365171.50/0/0
2026-10-22R802:12SAFETY TEST over budget 280/70/30 (WATCH_BELL_RATE_FIGHT=30, WATCH_BELL_RATE_REBUILD=280, WATCH_BELL_RATE_SCANNER=70)183 s94 s89 sno22 % / 16 %382174.00/0/0
2026-10-22R902:19CANDIDATE pump 225/40/20 (WATCH_BELL_RATE_FIGHT=20, WATCH_BELL_RATE_REBUILD=225, WATCH_BELL_RATE_SCANNER=40)86 s109 s67 sno0 % / 0 %338170.80/0/0
2026-10-22R1002:25CANDIDATE pump 225/40/20 (WATCH_BELL_RATE_FIGHT=20, WATCH_BELL_RATE_REBUILD=225, WATCH_BELL_RATE_SCANNER=40)81 s104 s71 sno0 % / 0 %322242.80/0/0
2026-10-22R1102:31candidate, one-fetcher OFF (WATCH_BELL_RATE_FIGHT=20, WATCH_BELL_RATE_REBUILD=225, WATCH_BELL_RATE_SCANNER=40)86 s121 s78 sno0 % / 0 %315173.90/0/0
2026-10-22R1201:23v2 fixes default119 s121 s-no0 % / 0 %287157.20/0/0
2026-10-22R1301:30v2 fixes default (repeat)116 s122 s-no0 % / 0 %287157.50/0/0
2026-10-22R1401:36v2 fixes, split 205/40/40 (WATCH_BELL_RATE_FIGHT=40, WATCH_BELL_RATE_REBUILD=205, WATCH_BELL_RATE_SCANNER=40)122 s125 s-no0 % / 0 %287181.40/0/0
2026-10-22R1501:43v2 fixes + FIGHT prefetch, split 205/40/40 (WATCH_BELL_RATE_FIGHT=40, WATCH_BELL_RATE_REBUILD=205, WATCH_BELL_RATE_SCANNER=40)128 s125 s-no0 % / 0 %289327.90/0/0
2026-10-22R1601:50ship config: v2 fixes + FIGHT prefetch + whole rule, 225/40/20129 s130 s-no0 % / 0 %288165.80/0/0
2026-10-22R1701:57ship config: v2 fixes + whole rule, prefetch OFF, 225/40/20122 s120 s205 sno0 % / 0 %289193.40/0/0
2026-10-22R1802:03ship config repeat (prefetch OFF, 225/40/20)116 s123 s221 sno0 % / 0 %288160.20/0/0
2026-10-22R1910:38[off-hours] FIGHT bell windows + Rolls wait + anchoring, plumbing--168 sno- / -157149.70/0/0
2026-10-22R2023:13items 2+3 merged (57ce9a1b): Rolls re-plan windows, FIGHT non-blocking verify incl. bell121 s85 s58 sno0 % / 1 %292148.50/0/0

How each tab is graded: from 2026-10-22 every tab on everything done (every row re-quoted after the bell) within 2 min (owner 2026-10-22). Before that (60 s): Rolls and Scanner on the rows the Desk shows, FIGHT on its bell check of every fortress, Coverage and Income on every row. A window miss is a contract asked for that returned no quote before its window closed. Goal A (actionable rows within 5 min, graded at 22:15 SGT): Open Reliability. Updated 2026-11-03 23:50 SGT.