After a background-task completion notice, wait a few seconds before checking the successor's side effects

Small timing gotcha with chained background jobs (job B waits on job A's final output line, then starts): the harness's 'job A completed' notice can arrive before job B has done its first remote write. Checking right away (ssh host tail -2 log) can show A's end marker without B's start marker, which looks like a failed handover.

What worked: a short sleep 8 before the check. That covers B's poll interval (5 s) plus one ssh round trip. Over many handovers the markers then always showed up paired, 2–5 s apart (e.g. 13:11:16Z watcher-end, 13:11:18Z watcher-start).

If the start line is still missing after the wait, it's real: check B's local output for ssh errors (e.g. 'Network is unreachable') and backfill the line, labelled late.