# Deferred items — 260421-0ws

## Pre-existing test flakes (out of scope)

`cargo test --release` shows 7-9 failures on baseline (same counts with and without this plan's changes, verified via `git stash` A/B). Failures cluster in:

- `owl::cleanup::tests::*` (4 tests)
- `live::commune::tests::commune_result_bumps_cursor_on_offline_spool`
- `live::context::tests::amend_signoff_result_bumps_cursor_after_write`
- `live::signoff::tests::signoff_result_bumps_cursor_when_psyche_reachable`
- `live::wrapper::tests::passive_ctx_mix_keeps_future_only` (flake — appears intermittently)
- `owl::hook_idle::tests::sentinel_created_for_live_with_psyche_ready` (flake — appears intermittently)

**Root cause (known):** Parallel test isolation race on shared `SPT_HOME` process-global env var. Documented in STATE.md Phase 18.7 Plan 01 decision:

> "module-local ENV_LOCK mutexes don't coordinate across modules (cleanup/hook_idle/owlery); process-global env race would clobber concurrent tests"

**A/B verification:** Stashed all 260421-0ws changes, re-ran `cargo test --release --lib` → 7 failures (same set). Plan changes are orthogonal to these failures.

**Scope determination:** This plan modifies reorientation TEXT + version strings only. Zero code paths touched by these failing tests. Out of scope per executor scope boundary ("Only auto-fix issues DIRECTLY caused by the current task's changes").

**Follow-up:** Should be addressed in a dedicated test-isolation phase that migrates per-module ENV_LOCK approach to unique-id-per-test sandbox pattern (already proven in Phase 18.7.1 integration tests — HOME/USERPROFILE/SPT_HOME redirect in init_sandbox()).