In T329064, the mediawiki.page-undelete stream was empty for over 3 months. No one noticed.
We should have some kind of automated throughput monitoring for important streams.
- Add some info to stream config indicating the approximate expected throughput of a stream
- Add a setting for enabling throughput alerting
- Write a script that can be executed by AlertManager or Icinga to check throughput of all streams with throughput alerting enabled.