flake: TestPGCoordinatorSingle_MissedHeartbeats_NoDrop

[flake-investigator]

CI Failure Details

CI Run Link: https://github.com/coder/coder/actions/runs/21424874845

Failed Job: test-go-pg (macos-latest) (Job ID: 61691881600)

Commit Info:

• SHA: 7090a1e205feaaf703e912058a4bf1fac6a9a389
• Author: Spike Curtis
• Date: 2026-01-28

Failing Test

• Package: enterprise/tailnet
• Test: TestPGCoordinatorSingle_MissedHeartbeats_NoDrop
• Source: enterprise/tailnet/pgcoord_test.go (function starts around line ~319)

Error / Symptoms

From the job log:

=== FAIL: enterprise/tailnet TestPGCoordinatorSingle_MissedHeartbeats_NoDrop (71.74s)
    pgcoord_test.go:353:
        Error Trace: /Users/runner/work/coder/coder/tailnet/test/peer.go:175
        Error:       Received unexpected error:
                    context deadline exceeded
        Test:        TestPGCoordinatorSingle_MissedHeartbeats_NoDrop
    pgcoord_test.go:357:
        Error Trace: /Users/runner/work/coder/coder/tailnet/test/peer.go:227
        Error:       Received unexpected error:
                    context deadline exceeded
        Test:        TestPGCoordinatorSingle_MissedHeartbeats_NoDrop
    pgcoord_test.go:361:
        Error:       Condition never satisfied
        Test:        TestPGCoordinatorSingle_MissedHeartbeats_NoDrop
    pgcoord_test.go:362: timeout waiting for responses to close for client

Notes / Suspected Root Cause

This appears to be a timing-sensitive flake in the Postgres-backed tailnet coordinator tests (missed heartbeat / LOST update behavior). The failure mode is consistent with the test not observing expected peer updates within testutil.WaitSuperLong.

Assignment Analysis

Intended git blame target:

git blame -L 319,375 enterprise/tailnet/pgcoord_test.go

I can’t run git blame directly with current bot tooling, so I used recent file history as a proxy:

• git log -- enterprise/tailnet/pgcoord_test.go (via API) shows repeated ownership/maintenance by Spike Curtis across multiple substantive changes.

Given that, assigning to Spike Curtis as the most likely owner/maintainer of this PG coordinator test area.

Related Issues / Duplicate Search

• Searched for TestPGCoordinatorSingle_MissedHeartbeats_NoDrop / MissedHeartbeats in coder/internal: none found.
• Similar (but different test) example: flake: TestPGCoordinatorSingle_AgentWithClient - timeout updating node for reconnection (PostgreSQL 17) #1045