OSDMap updates from pgmap can be delayed indefinitely
Like #9321, but in the other direction: we try to update the osdmap from the pgmap in OSDMonitor::tick(), but if pgmonitor->is_readable() fails, we just ignore it. As with the other bug, in our current implementation we should be safe to read the pgmap at any time because it's always consistent, even if we know it to be stale.
Beyond that, there is no output at any debug level if we fail to do the work, and we don't schedule a Context to retry. If this remains conditional we need to do something to make sure the work happens eventually, and add debug output at an appropriate level so we can trace actual activity as it happens.