flake: TestProvisionerDaemon_PSK (multiple variants)

[flake-investigator]

CI Run: https://github.com/coder/coder/actions/runs/24261779307
Job: test-go-pg-17
Commit: b14943313829b76b280703151688a683eb93d3b3 (Kayla はな)

Failing tests:

• TestProvisionerDaemon_PSK/OK
• TestProvisionerDaemon_PSK/AnotherOrgByNameWithUser
• TestProvisionerDaemon_PSK/NoUserNoPSK

Failure excerpt:

clitest.go:299: command "coder provisionerd start" exited with error: running command "coder provisionerd start": shutdown:
    github.com/coder/coder/v2/enterprise/cli.(*RootCmd).provisionerDaemonStart.func1
        /home/runner/work/coder/coder/enterprise/cli/provisionerdaemonstart.go:267
  - context canceled

coderd: GET ... /api/v2/organizations/default/provisionerdaemons/serve ... status_code=500 ...
response_body= {
    "message": "Internal error fetching organization \"default\".",
    "detail": "fetch object: context canceled"
}

Error analysis:

• Provisioner daemon CLI starts, then immediately attempts graceful shutdown and exits with context cancellation.
• coderd endpoints for provisioner daemons return 500 due to context canceled while the test expects the daemon to keep running.
• Likely timing/teardown race in the TestProvisionerDaemon_PSK subtests (parallel tests + short-lived contexts).

Root cause classification: Flaky test (timing/teardown/context cancellation)

Precise assignment analysis:

• git log --oneline -10 --follow enterprise/cli/provisionerdaemonstart_test.go
  • 4d1003e fix: remove initial global HTTP client usage (Zach)
  • Next most recent change is much older; assigning to most recent meaningful modifier of this test file.

Race/Panic/OOM evidence:

• No WARNING: DATA RACE, panic, or OOM indicators found in the failing job logs.

Reproduction:

• go test ./enterprise/cli -run TestProvisionerDaemon_PSK -count=1