Having just witnessed the same symptoms on our production service I performed a quick investigation.
This revealed that dhcpd was iterating over a gettimeofday ... select loop in which select was returning immediately with three sockets to the failover peer that were in CLOSE_WAIT state. I was unable to debug further since restoring service was a priority. I shall add this to my ever growing list of failure modes for the ISC DHCP server failover implementation. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org