FE-1261: Recover prepared execution journals

[kostandinang]

Stack context

Follow-up to #371 under FE-1261. The original PR merged while its latest automated review was still reporting.

Why

drive() prepared Petri observation on disk but immediately reconciled with the stale pre-preparation metadata object. A run advanced by a direct lifecycle tool could therefore halt with petri_journal_gap instead of repairing the journal.

What

• Refresh run.json after observation preparation before lifecycle reconciliation.
• Cover an unprepared run advanced directly to worktree_created, proving reconciliation and forward progress.

Verification

• Affected executor suite: 272/272 passing.
• Production build and type check passing.

[cursor]

PR Summary

Medium Risk
Touches the core drive() reconciliation path in the executor; scope is narrow (metadata refresh after preparation) but incorrect behavior could still affect run progression and journal repair.

Overview
Fixes a bug where drive() prepared Petri observation on disk but still reconciled lifecycle steps against the stale in-memory run.json snapshot from the start of the loop iteration.

After preparePetriObservation, the drive loop now re-reads run.json and updates state before reconcilePreparedLifecycleJournal. That way runs advanced by direct lifecycle tools (e.g. worktree already created while metadata still looked unprepared) reconcile against current provenance instead of halting with petri_journal_gap.

Adds an orchestrate test for an unprepared run with a worktree already in place, asserting reconciliation fires worktree_create and populate and reaches worktree_populated with petriObservationPrepared: true.

^{Reviewed by Cursor Bugbot for commit 2561716. Bugbot is set up for automated code reviews on this repo. Configure here.}

[kostandinang]

• FE-1261: Recover prepared execution journals #373 👈 (View in Graphite)
• next

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