FE-1261: Recover prepared execution journals

[kostandinang]

Stack context

FE-1261 is a repair beneath FE-1211's comparison-execution work.

Why

Direct lifecycle tools and superseding runs could leave run.json ahead of the prepared Petri journal, halting resumable comparison execution or misclassifying journal authority.

What

• Reconcile causally valid trailing lifecycle facts without repeating completed effects.
• Fail closed on unreadable or divergent journals while preserving valid parallel interleavings.
• Prepare complete Petrinaut observation artifacts for superseding runs and report standalone journal failures accurately.

How to test

• npm run test -- src
• npm run build
• npm run check:markdown-links
• npm run check:skills
• npm run check:promoted-run-paths

[kostandinang]

• FE-1261: Recover prepared execution journals #371 👈 (View in Graphite)
• next

This stack of pull requests is managed by Graphite. Learn more about stacking.

[cursor]

PR Summary

High Risk
Changes crash-durable Petri journal/marking authority, drive halt/resume semantics, and superseding run creation—core executor correctness paths where mistakes can corrupt or mis-classify run state.

Overview
Prepared runs can get run.json ahead of the Petri journal when lifecycle steps run outside drive() or after supersession. This PR adds reconcilePreparedLifecycleJournal to catch up a causally valid serial prefix (through report_init or a standalone slice_start) by appending missing transition events and refreshing marking.json, without re-running completed effects.

initializeReports, startSlice, and drive() call reconciliation before dispatch and around report_init / slice_start so those steps no longer double-append journal events. Blocked outcomes surface concrete reasons (petri_journal_gap, petri_input_unreadable, petri_marking_persist_failed, terminals, active parallel batch) instead of lumping many cases into parallel_batch_active.

createSupersedingRun now preparePetriObservation under the target run’s execution lease (rollback on failure) and documents the extra petrinaut side effects, matching execute_run_create.

^{Reviewed by Cursor Bugbot for commit 2d2c774. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 2d2c774. Configure here.}

[cursor]

Stale state skips journal repair

Medium Severity

After preparePetriObservation persists petriObservationPrepared, drive still passes the earlier in-memory state into reconcilePreparedLifecycleJournal. That helper treats the run as not_prepared and skips catch-up, so a lifecycle-ahead journal left by direct steps can still halt as petri_journal_gap instead of being repaired.

Additional Locations (1)

• src/executor/petri-lifecycle-reconciliation.ts#L40-L41

 

^{Reviewed by Cursor Bugbot for commit 2d2c774. Configure here.}