Also, we are getting our pods evicted due to storage limits, caused by 21G

/tmp/osquery_status

files. My gut says this is related to the lack of database connections (like, osquery_status gets cleared when it gets written to the database, or something). Is that correct?

This is not related to lack of DB connections. You're writing the osquery status logs to disk (because you haven't configured anything else) and it sounds like that is filling up the disk. Maybe you want to use the built-in log rotation? https://github.com/fleetdm/fleet/blob/master/docs/infrastructure/configuring-the-fleet-binary.md#filesystem_enable_log_rotation

d'oh, yes, how'd I miss that?