Skip to content

fix(reconciler): P0+P1 bug fixes from audit #1986 (v3) #2031

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
jrf0110 wants to merge 2 commits into from
Closed

fix(reconciler): P0+P1 bug fixes from audit #1986 (v3) #2031

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4bd4fea
fix(reconciler): parameterize circuit breaker query, remove duplicate…
Apr 5, 2026
a70365d
fix(reconciler): use last_poll_at for Rule 4 staleness check instead …
Apr 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 12 additions & 6 deletions cloudflare-gastown/src/dos/town/reconciler.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,7 @@ const CIRCUIT_BREAKER_WINDOW_MINUTES = 30;
* beads that eventually succeeded (status = 'closed').
*/
function checkDispatchCircuitBreaker(sql: SqlStorage): Action[] {
const cutoff = new Date(Date.now() - CIRCUIT_BREAKER_WINDOW_MINUTES * 60_000).toISOString();
const rows = z
.object({ failure_count: z.number() })
.array()
Expand All @@ -61,11 +62,11 @@ function checkDispatchCircuitBreaker(sql: SqlStorage): Action[] {
/* sql */ `
SELECT count(*) as failure_count
FROM ${beads}
WHERE ${beads.last_dispatch_attempt_at} > strftime('%Y-%m-%dT%H:%M:%fZ', 'now', '-${CIRCUIT_BREAKER_WINDOW_MINUTES} minutes')
WHERE ${beads.last_dispatch_attempt_at} > ?
AND ${beads.dispatch_attempts} > 0
AND ${beads.status} != 'closed'
`,
[]
[cutoff]
),
]);

Expand Down Expand Up @@ -1049,8 +1050,7 @@ export function reconcileReviewQueue(
b.${beads.columns.rig_id}, b.${beads.columns.updated_at},
b.${beads.columns.metadata},
rm.${review_metadata.columns.pr_url},
b.${beads.columns.assignee_agent_bead_id},
b.${beads.columns.metadata}
b.${beads.columns.assignee_agent_bead_id}
FROM ${beads} b
INNER JOIN ${review_metadata} rm ON rm.${review_metadata.columns.bead_id} = b.${beads.columns.bead_id}
WHERE b.${beads.columns.type} = 'merge_request'
Expand Down Expand Up @@ -1140,12 +1140,18 @@ export function reconcileReviewQueue(
// Rule 4: PR-strategy MR beads orphaned (refinery dispatched then died, stale >30min)
// Only in_progress — open beads are just waiting for the refinery to pop them.
// Skip when refinery code review is disabled: poll_pr keeps the bead alive via
// updated_at touches, and no refinery is expected to be working on it.
// last_poll_at touches, and no refinery is expected to be working on it.
// Use last_poll_at (set by poll_pr each tick) to determine staleness; fall back
// to updated_at for old beads that predate the last_poll_at field.
if (
refineryCodeReview &&
mr.status === 'in_progress' &&
mr.pr_url &&
staleMs(mr.updated_at, ORPHANED_PR_REVIEW_TIMEOUT_MS)
staleMs(
(typeof mr.metadata?.last_poll_at === 'string' ? mr.metadata.last_poll_at : null) ??
mr.updated_at,
ORPHANED_PR_REVIEW_TIMEOUT_MS
)
) {
const workingAgent = hasWorkingAgentHooked(sql, mr.bead_id);
if (!workingAgent) {
Expand Down