Large repositories exceed fixed worktree creation and status-probe timeouts

[chubes4]

Problem

Data Machine Code cannot create or resolve a worktree for the WPCOM repository within its fixed process budgets, even when the primary is fresh and the requested base is a local origin/trunk ref.

Two failures were reproduced:

Worktree creation

studio wp datamachine-code workspace worktree add \
  wpcom fix/dotprod-160-disable-chatgpt-site-builder \
  --from=origin/trunk \
  --skip-bootstrap \
  --skip-context-injection \
  --format=json

Result:

Process command timed out after 30 second(s).

No worktree or branch remained. Native Git then materialized the same 223,451-file worktree successfully in about 40 seconds:

git worktree add -b fix/dotprod-160-disable-chatgpt-site-builder \
  /Users/chubes/Developer/wpcom@fix-dotprod-160-disable-chatgpt-site-builder \
  origin/trunk

Targeted lookup

After native creation and successful DMC refresh-context, exact lookup failed because its status probe has a five-second budget:

worktree_get_status_probe_failed
Targeted worktree lookup status probe failed: Process command timed out after 5 second(s).
cause_code: git_command_timeout

Native git status --short --branch completed and showed a clean branch tracking origin/trunk.

Impact

• DMC's supported creation path cannot manage a repository of this size.
• Homeboy's DMC provider cannot resolve the resulting exact handle, blocking Cook before dispatch.
• The operator had to use native git worktree add and then re-adopt the checkout through refresh-context.
• Fixed budgets turn repository size into a correctness failure instead of a bounded slow operation with progress.

Expected contract

Worktree lifecycle operations should have operation-appropriate budgets and preserve durable progress for large repositories. Cheap inventory probes can stay short; materialization and exact safety probes need distinct timeout policy.

Acceptance criteria

• Give worktree materialization a configurable or repository-scaled timeout greater than the current fixed 30 seconds.
• Emit progress/heartbeat evidence while Git checks out a large worktree.
• Give exact git status safety probes a configurable budget or a bounded alternate probe suitable for large repositories.
• Preserve fail-closed cleanliness semantics when status cannot be verified.
• Return a typed timeout with a resume/retry command and any partial-resource cleanup evidence.
• Test a synthetic large repository where checkout exceeds 30 seconds and status exceeds five seconds.
• Prove create, get, refresh-context, Homeboy provider resolution, and cleanup operate without requiring native Git adoption.

AI assistance

• Model: OpenAI GPT-5.6 Sol
• Tool: OpenCode
• Used for: Reproduced the DMC creation and exact-lookup timeouts, compared them with successful native Git behavior, searched existing Data Machine Code issues, and drafted this report. Chris Huber reviewed and owns the issue.