perf(planning): accept the speed target and lower the hang-guard timeout #286

New Issue

2026-10-16T22:39:46+02:00

nessar commented

2026-10-16 22:39:46 +02:00

Parent

What to build

Confirm the speed target and close out the performance work. The loaded year finishes under 10 seconds and every other scenario under 3 seconds, with the existing protocol in Node on the development machine. The relative gate's multipliers are tightened to the target, and the outer timeout drops from 300 seconds to about three times the expected run, as a hang guard only.

Acceptance criteria

With the existing protocol (one warm-up, two measured repetitions, 2 GiB heap, one worker), the loaded year is under 10 seconds and every other scenario under 3 seconds, recorded in the benchmark documentation.
The relative gate's multipliers match the target. The outer timeout is lowered to about three times the measured expected run, and the CI documentation is updated.
The CI benchmark passes on at least two runners with the lowered timeout.
If the target is not reached with exact optimizations, record the remaining gap and the profile breakdown, and leave this ticket open for a separate decision on search changes.
Note in the benchmark documentation whether ADR 0003's five-second target should be revisited for the loaded year under composable rules.

Blocked by

## Parent #279 ## What to build Confirm the speed target and close out the performance work. The loaded year finishes under 10 seconds and every other scenario under 3 seconds, with the existing protocol in Node on the development machine. The relative gate's multipliers are tightened to the target, and the outer timeout drops from 300 seconds to about three times the expected run, as a hang guard only. ## Acceptance criteria - [ ] With the existing protocol (one warm-up, two measured repetitions, 2 GiB heap, one worker), the loaded year is under 10 seconds and every other scenario under 3 seconds, recorded in the benchmark documentation. - [ ] The relative gate's multipliers match the target. The outer timeout is lowered to about three times the measured expected run, and the CI documentation is updated. - [ ] The CI benchmark passes on at least two runners with the lowered timeout. - [ ] If the target is not reached with exact optimizations, record the remaining gap and the profile breakdown, and leave this ticket open for a separate decision on search changes. - [ ] Note in the benchmark documentation whether ADR 0003's five-second target should be revisited for the loaded year under composable rules. ## Blocked by - #284 - #285

nessar added the ready-for-agent label 2026-10-16 22:39:46 +02:00

nessar added a new dependency 2026-10-16 22:39:46 +02:00

#284 perf(planning): prune candidates that cannot beat the protected prefix

nessar added a new dependency 2026-10-16 22:39:46 +02:00

#285 ci(benchmark): gate scenario speed relative to the runner

nessar referenced this issue from a commit

2026-10-17 00:52:35 +02:00

ci(benchmark): gate scenario speed relative to the runner

nessar referenced this issue from a commit

2026-10-17 00:52:35 +02:00

docs(benchmark): record the remaining gap to the speed target

nessar referenced this issue