Comment 4 for bug 1822096

#1 is triggered via (it seems a guest goes away - qemu died and then cleanup runs after a write failed)
 daemonStreamHandleWrite (failing to write)
 -> daemonStreamHandleAbort (closing things and cleaning up)
    -> ... virChrdevFDStreamCloseCb
        virMutexLock(&priv->devs->lock);
        # priv = Domain's device information structure
        # client stream is client=0x2aa465d8f60
        # there would be code that is meant to avoid issues "Unable to close" if lock held
        # but the log doesn't show this

#10 seems triggered via an "undefine" call
  remoteDispatchDomainUndefine
  ... -> virChrdevFree
     ... -> virFDStreamSetInternalCloseCb
        -> virObjectLock(virFDStreamDataPtr fdst)
          -> virMutexLock(&obj->lock);
  #closing all streams of a guest (requiring the same locks)