The integration reports that it is alive

A process that syncs in the background has a worse failure mode than going down: staying up and doing nothing. Nobody finds out until a booking is missing.

Every sync reports in, and there is a status endpoint that can be watched from outside. If one specific booking fails, it retries on its own by its identifier rather than redoing the whole batch.

This is also what makes a real incident page possible, the day there is one: with no signal to watch, a status page is decoration.