`Close()` has a redundant `time.Sleep(1s)` — adds fixed latency to every teardown

[never-agent]

Summary

Close() ends with time.Sleep(time.Second) to "wait for process to fully terminate", but Stop() already calls a.cmd.Wait(), which blocks until the process is reaped. The sleep is redundant and adds a fixed 1s to every teardown.

Detail

anvil.go (Close):

if stopErr := a.Stop(); stopErr != nil { ... }
// Wait for process to fully terminate
time.Sleep(time.Second)

Stop() already does _ = a.cmd.Wait() after Process.Kill(), which is the real synchronization point — by the time Stop() returns, the process is reaped. The trailing sleep therefore adds 1s of pure wall-clock to every Close(). Across a suite that spins instances up and down (or even the shared-anvil teardown), that is measurable dead time.

Suggested fix

Remove the time.Sleep. If there's a residual concern about port release before the next instance binds the same port, poll the port for closure (bounded) rather than sleeping a fixed second.

Confidence

High — read directly from source. Wait() semantics are well-defined; the sleep adds no synchronization Wait() doesn't already provide.

Found during an external read-through of go-anvil@3701910. Not covered by any open v0.2.0 issue.