Ceph » rgw

the RGWIndexCompletionThread stores a list of index completion ops that failed with ERR_BUSY_RESHARDING, for synchronous retry under guard_reshard() in RGWIndexCompletionThread::process()

however, after process() finishes it's work, it goes to sleep on its condition variable without checking for more completions. so it's easy for RGWIndexCompletionThread to miss signals on this condition variable, and not wake up even though there's still work to do. once completions are stalled this way, nothing will wake up the thread again until new completions race with another reshard