Skip to content

Block 89a: design/89 probes P1–P4 (no product code) - #60

Merged
zacsimile merged 20 commits into
mainfrom
block-89a
Oct 19, 2026
Merged

zacsimile merged 20 commits into
mainfrom
block-89a

Conversation

@zacsimile

Copy link
Copy Markdown
Member

Block 89a of design/89: standalone probes, no product code.

  • design/89-probes.py with subcommands p1..p4 plus --selftest (bridge-shaped fakes written from pycro-manager/ndstorage source; mutation controls for every verdict). It imports nothing from microclaw.
  • One-line launchers design/89-probe-p{1..4}.ps1 (installed slot interpreter, -I, no property name in any command).
  • Runbook design/89-block89a-demo-probes.md (demo machine; implementation pinned at 90fdf18).
  • R157 updated: the slow state is transient and cleared without a reboot. Gap p95 separates the two states (12.03–12.05 ms healthy, n=8; 21.56–22.33 ms slow, n=4). The cadence probes carry interleaved controls and report NOT EXERCISED in the slow state (operator decision, 2026-10-18).

Review (step 3): two revision rounds. Round 1 fixed:

  • a notification callback that could kill pycro-manager's dispatcher;
  • the burst witness, moved from a guessed CoreLog regex to the engine's camera notifications;
  • a missing plain-burst control in P3, with both refutation shapes scored as FAIL;
  • a P3 metadata precision mismatch;
  • probe-side index polling that contaminated the P1/P2/P4 arms;
  • a gap in the R157 guard over P4's fast block;
  • report size.

Round 2 added a disk preflight and cleanup of scored image stacks. Coordinator suite run: 5963 passed, 107 skipped. Selftest: SELFTEST PASSED.

Owed after the demo run: rig results, design reconciliation, and fact 5 into CLAUDE.md if P3 confirms it. The M2 limbs of P1/P4 are owed, unbooked.

…ates; 89a cadence probes carry a control arm (operator decision)
…lied inside a burst (design/89 P3, demo, 2026-10-19); R157: slow again 2026-10-19 on the pycro-manager route
…un episode not locked to writes; bridge round trip ~5x slower in the slow state
…ram and MM uptime are the candidates, still confounded
…icro-Manager between the slow and healthy runs
…n; the operator restarted Micro-Manager before the probes; the snapshot StartTime is unreconciled
…owth FAILs recorded), P2/P3 results, preallocated Java index for F3; R157 restart observation; R159 M2 limbs owed
@zacsimile
zacsimile merged commit 0ffa44a into main Oct 19, 2026
2 checks passed
@zacsimile
zacsimile deleted the block-89a branch October 19, 2026 14:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant