flake: TestAPI/Delete/OK_with_container_and_subagent

[flake-investigator]

CI Failure Details

• CI Run: https://github.com/coder/coder/actions/runs/21444217040
• Failed Job: https://github.com/coder/coder/actions/runs/21444217040/job/61755356642 (test-go-pg / ubuntu-latest)
• Branch: main
• Commit: e13f2a986901d14d6106d115883a128accf659b1 (author: Steven Masley)
• Date: 2026-01-28 15:32:18Z

Root Cause Classification

• Type: Flaky Test (mock expectation + timeout)
• Affected package: agent/agentcontainers
• Failing test: TestAPI/Delete/OK_with_container_and_subagent

Error Evidence

=== FAIL: agent/agentcontainers TestAPI/Delete/OK_with_container_and_subagent (25.01s)
api.go:1783: Unexpected call to *acmock.MockContainerCLI.DetectArchitecture(...)
  expected call at api_test.go:364 has already been called the max number of times
api_test.go:1481: refresh containers failed:
    github.com/coder/coder/v2/agent/agentcontainers.(*API).RefreshContainers.func1
    /home/runner/work/coder/coder/agent/agentcontainers/api.go:1168
  - context deadline exceeded
controller.go:97: missing call(s) to *acmock.MockContainerCLI.Stop(...)
controller.go:97: missing call(s) to *acmock.MockContainerCLI.Remove(...)

Analysis

• The subtest starts the updater loop and injects a subagent, then explicitly calls api.RefreshContainers(ctx) before issuing the DELETE request.
• During refresh, a background updater invocation triggers an extra DetectArchitecture call, exceeding the gomock expectation and leading to a context deadline timeout.
• The timeout short-circuits the delete flow, so the expected Stop/Remove calls are never observed.
• No panic/OOM or race detector output was observed in this job.

Ownership / Assignment

Primary assignment uses test function blame per guidelines.

Relevant test lines:

• File: agent/agentcontainers/api_test.go
• Subtest: "OK with container and subagent" around line ~1363 (log references api_test.go:1406/1481)

Suggested blame commands:

grep -n "OK with container and subagent" agent/agentcontainers/api_test.go
# -> line ~1363

git blame -L 1363,1460 agent/agentcontainers/api_test.go

Recent change context:

• list_commits for the file shows commit 44a46db487 ("feat(agent): support deleting dev containers") by Danielle Maywood as the most recent substantive change to delete devcontainer tests; newer commits are import-only.

Assigning to Danielle Maywood for triage.

Related Issues

• flake: agent/agentcontainers TestAPI/Error/DuringInjection #1194 (different test: TestAPI/Error/DuringInjection)
• flake: Data race in agent/agentcontainers - TestAPI/Error/DuringUpWithContainerID #1169 (data race in TestAPI/Error/DuringUpWithContainerID)

Reproduction

go test ./agent/agentcontainers -run TestAPI/Delete/OK_with_container_and_subagent -count=50