Observal hook integration with Copilot CLI is broken end‑to‑end after CLI upgrade: stale module reference, fail‑closed timeout, and doctor doesn't detect it

[Lokesh7025]

Checked for duplicates?

• This issue is not a duplicate

What are the steps to reproduce this bug?

1. Install observal-cli on a machine that previously had an older version that wrote Copilot CLI hooks (or run observal install then upgrade the CLI past the copilot_cli_hook →
   copilot_cli_session_push rename).
2. Launch copilot from any directory and run any tool, e.g. gh issue list.
3. Observe in ~/.copilot/logs/process-*.log:
   - No module named observal_cli.hooks.copilot_cli_hook (from ~/.copilot/settings.json entries).
   - preToolUse hook from "/home//.copilot/hooks/observal.json" execution failed (fail-closed): Hook command timed out after 5 seconds.
   - Tool calls denied with Denied by preToolUse hook from "/.../observal.json".
4. Run observal doctor, observal doctor patch --all --harness copilot-cli --dry-run, and observal doctor cleanup --harness copilot-cli --dry-run.
   - doctor reports "All clear".
   - patch reports "Already up to date".
   - cleanup only lists ~/.copilot/hooks/observal.json and never mentions the stale settings.json entries.

Expected behaviour

• CLI upgrades that rename hook modules must migrate existing references in ~/.copilot/settings.json (no orphaned copilot_cli_hook paths).
  • preToolUse telemetry hook should be fail‑open (or async/fire‑and‑forget) so slow/unreachable telemetry never blocks user tool calls; the 5 s timeout is too aggressive for a
    network call.
  • observal doctor should detect references to known‑removed module paths in any managed harness file (including settings.json) and offer to auto‑migrate them under --yes;
    cleanup should strip them too.

Support bundle

No response

(Optional) Anything else you want to share?

No response