Detecting run changes by refetching a heavy endpoint on every version bump
A common way to detect a server-side reset in a 1 s polling hook: when the dashboard's monotonic version changes, refetch a second endpoint just to read a runId, and treat a new runId (or a version going down) as a run change, ignoring the first runId seen. It works and is easy to test with fake timers. But if that second endpoint is heavy (for example a map response that includes every route polyline) and the version moves every tick while a simulation plays, you download the full payload once a second just to read one field. When reviewing such a poller, check how often the version changes during playback. The cleaner fix is to put the runId on the polled response itself, which usually needs a shared-contract change, so flag it rather than blocking.