docs: update context-PR + slice-BRC observability for #2684 [doc-updater]

docs/reference/orchestrator-cli.md

@@ -461,7 +461,7 @@ The orchestrator manages the Context PR end-to-end (create branch, commit refine
 
 **Observability:** If the context PR hook runs but does not open a PR, the wrapper at `_maybe_open_base_pr_for_plan_to_implement` surfaces a `context_pr.failed` event (the inner hook raised) or `context_pr.skipped` event (the inner hook returned without raising, but the post-hook contract still records `contract.pr.context_pr_number = null`) on three sinks (#2611): the in-process `MessageStore` (so the event shows up as a `CONTEXT_PR_FAILED` / `CONTEXT_PR_SKIPPED` entry in `recent_messages` and `/pipelines/<id>/messages`), the `EventBus` (so `/status/wait` long-pollers wake — both event types are in `_STATUS_WAIT_EVENT_TYPES`, and the message types are in `_STATUS_WAIT_MESSAGE_TYPES`), and the legacy `StatusReporter` handler chain (no production handler is registered today, but the call is preserved for any future console/file handler). The emit branch is gated on `pipeline.repo` and `pipeline.base_branch` both being truthy; local-mode pipelines (no remote, no base branch) skip the hook silently and produce neither event (#2593).
 
-The wrapper also writes structured log lines on every invocation: `Context PR hook entered (#2548)` on entry, followed by either a short-circuit log line from the inner hook (e.g. `Context PR hook: pipeline has no remote repo, skipping`) or the wrapper's `Context PR hook raised at plan→implement transition (continuing) (#2548)` warning when the inner hook raised. (CUSTOM-mode pipelines short-circuit in the wrapper before the inner hook runs and instead emit `Context PR hook skipped (CUSTOM mode) (#2548)`; this only matters for log forensics on stuck pipelines, since CUSTOM-mode pipelines are single-phase (#1762) and never traverse plan→implement in normal operation.) The `context_pr.skipped` case corresponds to an inner contract-side short-circuit that did not raise — for example, no `pr` block on the contract, the post-hook contract reload failed, or a late `save_contract` failure swallowed the success after the gateway already opened the PR on GitHub.
+The wrapper also writes structured log lines on every invocation: `Context PR hook entered (#2548)` on entry, followed by either a short-circuit log line from the inner hook (e.g. `Context PR hook: pipeline has no remote repo, skipping`) or the wrapper's `Context PR hook raised at plan→implement transition (continuing) (#2548)` warning when the inner hook raised. (CUSTOM-mode pipelines short-circuit in the wrapper before the inner hook runs and instead emit `Context PR hook skipped (CUSTOM mode) (#2548)`; this only matters for log forensics on stuck pipelines, since CUSTOM-mode pipelines are single-phase (#1762) and never traverse plan→implement in normal operation.) The `context_pr.skipped` case corresponds to an inner contract-side short-circuit that did not raise — for example, no `pr` block on the contract, the post-hook contract reload failed, or a late `save_contract` failure swallowed the success after the gateway already opened the PR on GitHub. A missing `WORKTREE_BASE_DIR` volume mount is another `context_pr.skipped` trigger: the hook falls back to `/tmp` (gateway-rejected path), logs `Context PR hook: WORKTREE_BASE_DIR missing — falling back to system temp (likely a broken volume mount in production…) (#2684)` at WARNING level, and the gateway silently rejects the push (#2684).
 
 In any of these cases, check `contract.pr.context_pr_number` (via `egg-contract show`) and the remote PR list before concluding the PR is missing — that last short-circuit can leave a PR open on GitHub while the contract still records `null`. **Pipeline deletion does not clean up Context PRs:** `egg-orch pipeline delete <id>` only removes the pipeline tip branch (`egg/<id>/work`) and per-container worktree branches; the Context PR branch (`egg/<id>/context`) is a sibling of the pipeline tip — same convention as the slice integration branches `egg/<id>/slice-N` — and is **not** deleted (see `_cleanup_remote_branches` in `orchestrator/routes/pipelines.py`). To remove a Context PR opened by an unwanted run the operator must close the PR and delete the branch manually:
 
@@ -470,7 +470,7 @@ gh pr close <context_pr_number>
 git push origin --delete egg/<pipeline-id>/context  # (gateway-mediated push)
 ```
 
-Per-slice implement-phase BRC history files written by the orchestrator (`.egg-state/brc-history/<id>-implement-slice-<N>.{md,json}` plus `<id>-implement-unattributed.{md,json}`) are visible in each slice PR's diff; the aggregate `<id>-implement.{md,json}` file is **not** produced in slice-aware mode. Babysit-pr and other non-slice runs continue to emit the single content-addressed file (`pr-<N>-<short-sha>-implement.{md,json}` for babysit). See [Orchestrator Architecture: BRC-history file naming](../architecture/orchestrator.md#brc-history-file-naming) for the full file-pattern table.
+Per-slice implement-phase BRC history files written by the orchestrator (`.egg-state/brc-history/<id>-implement-slice-<N>.{md,json}` plus `<id>-implement-unattributed.{md,json}`) are visible in each slice PR's diff; the aggregate `<id>-implement.{md,json}` file is **not** produced in slice-aware mode. Babysit-pr and other non-slice runs continue to emit the single content-addressed file (`pr-<N>-<short-sha>-implement.{md,json}` for babysit). See [Orchestrator Architecture: BRC-history file naming](../architecture/orchestrator.md#brc-history-file-naming) for the full file-pattern table. If the BRC history file is absent from a slice PR's diff, check orchestrator logs for `Per-slice BRC commit: WORKTREE_BASE_DIR missing — falling back to system temp…` — this WARNING indicates the `/home/egg/.egg-worktrees` volume is not mounted on the orchestrator pod, causing the gateway to reject the push (#2684).
 
 ## Related CLIs