For instance, we want to resend the latency query, in case the upstream latency has changed.
Probably needs !356 first. Otherwise, this scenario breaks:
- fallbackswitch is running
- something upstream of the fallbackswitch sinkpad is unlinked
- fallbackswitch receives reconfigure event and sends latency query upstream
- latency query fails because upstream is unlinked
- aggregator in fallbackswitch waits forever for the latency query, never switches to fallback