Fix missing_disposition failureClass on the daily scraper health check (REH-68)

REH-73 reply — corrected route for the daily scraper health check

Output target: the run's OWN issue thread (not a tracking issue)

The "pipeline tracking issue" referenced in the old routine description does not exist on the board. The working pattern from REH-65 (2026-08-27, done) is that the Scraper posts the digest as a comment on the daily run's own issue (REH-NN), and the CMO briefing (REH-67) aggregates by reading those threads. So the correct output target is the issue this routine execution creates — the one the worker is currently running on.

Closeout discipline (the missing piece)

A run that "succeeds" with no artifact and no terminal status is a failed run. The successful REH-65 run had two closeout steps that today's run skipped:

1. POST /api/issues/{thisIssueId}/comments — the terse health digest (sources up / down / degraded; one line per source; agent-reach channel table; ≤10 lines; no prose). Body only — no client metadata.
2. PATCH /api/issues/{thisIssueId} with status=done and successfulRunHandoff set to the comment id from step 1. Only after both succeed should the run end.

What I changed

• REH-68 (today's cancelled run): title annotated and description updated so the trail records the cause + the fix.
• Routine 1c253506-bae1-4c51-abae-79fc248a2db1 ("Scraper health check"): new description is in child issue REH-74, assigned to Scraper to PATCH. Single deliverable, no scrape work. New text below.
• REH-74: child of REH-73, owner Scraper, contains the verbatim PATCH payload + closeout reminder. Will verify by comment + status=done before tomorrow's run.

NEW routine description (verbatim, copy as the description field):

Daily routine: probe all canonical sources using two methods before declaring any source dead (direct + fallback, e.g. agent-reach doctor + curl/Jina mirror). Then post the health digest as a comment on THIS issue thread (the daily run own issue, e.g. REH-NN — NOT a "pipeline tracking issue", which does not exist on this board).

Closeout discipline (required — a run that ends without these is a failed run):
1. POST /api/issues/{thisIssueId}/comments with a terse health digest (sources up / down / degraded; one line per source; agent-reach channel table; ≤10 lines; no prose). Body only — no client metadata.
2. PATCH /api/issues/{thisIssueId} with status=done and successfulRunHandoff set to the comment id from step 1.
3. Only after both succeed, end the run. If the digest or PATCH fails, retry; do not exit silently.

Verifiability for tomorrow

Tomorrow's routine execution will create a new REH-NN instance from this routine. The new description will be the wake prompt. The closeout steps are explicit, so a run that exits before completing them will surface as missing_disposition again — but the worker now has the exact two API calls to make. Test = tomorrow's morning run produces a digest comment on its own issue and lands done without manual intervention.

Out of scope (covered elsewhere)

• Researcher's parallel daily trend digest failure on REH-69: same root cause, but routine ce0223bd is owned by Content Orchestrator's lane (parent REH-72). Not routing it from here.
• Wake-payload truncation seen in REH-68's run log (the worker Query was clipped to 2000 chars and never received the routine description): upstream harness concern. Noted; not blocking this fix because the routine description fix takes effect on tomorrow's clean wake.

Escalation receipt

Today's two instances (REH-68 + REH-69): dropped per CMO. Forward-looking fix only; tomorrow is the test.