Summary

Fixes #284 — tasks in the services view now refresh on the fly while expanded.

Root cause

refreshExpandedTasksCmd and checkServicesCmd are dispatched concurrently from the TickMsg handler (every 2s). checkServicesCmd calls TriggerRefreshIfNeeded() which starts an async (non-blocking) background goroutine to refresh the snapshot. Meanwhile, refreshExpandedTasksCmd calls GetTasksForService() → GetSnapshot() which returns the stale cached snapshot — the background refresh hasn't completed yet. On the next tick, the snapshot TTL (3s) often hasn't expired, so no new refresh is triggered. The result: tasks are perpetually read from stale data.

Collapsing and re-expanding works because the manual toggle happens at a different point in the refresh cycle — typically after the background refresh has completed — so GetSnapshot() returns fresh data.

Fix

Replace the tea.Batch of per-service closures with a single tea.Cmd that calls GetOrRefreshSnapshot() (blocking, synchronous) before fetching tasks. This guarantees the global snapshot cache is fresh when GetTasksForService() reads from it. The blocking call runs in a goroutine (tea.Cmd), so it does not freeze the UI. When the snapshot is already fresh (within 3s TTL), GetOrRefreshSnapshot() returns instantly from cache.

Flow after fix

TickMsg (every 2s)
  → checkServicesCmd goroutine (unchanged, async)
  → refreshExpandedTasksCmd goroutine:
      1. GetOrRefreshSnapshot()   ← blocks ~0.5s if stale, instant if fresh
      2. GetTasksForService(sid)  ← reads FRESH snapshot
      3. returns AllTasksLoadedMsg
  → Update() stores fresh tasks in m.serviceTasks
  → View() re-renders with updated task rows

Test plan

• Unit tests pass (go test ./views/services/...)
• Lint passes (golangci-lint run ./...)
• Build passes
• Manual: expand a service with p, scale/restart it, verify tasks update without collapsing

🤖 Generated with Claude Code