on the event that tsmon is unable to upload metrics, it logs a warning/error. If flushing runs in a tight loop, like it does in CQ, and there is a long period of time (hours) where it is impossible to send metrics, tsmon spams logs to the extent that it hits limits of other systems, e.g. logging system.
This caused https://irm.corp.google.com/incidents/a.fnj66Wxpx1YGDKxdvYQX/investigate

There is no value in logging the same message every 2sec. At the very lease intervals between such messages should be exponential.