From Ken Cormack on the MIMEDefang List:

Over the 6 years that our servers have been running, we've turned down (WAY
down) most, if not all of the various timeout values, to eliminate the
tendency for slow connectors to exhaust our confMAX_DAEMON_CHILDREN
connection pool.  I've always seen a fair number of log entries such as you
show, but we've had no complaints to date, and it's helped prevent a lot of
otherwise open-but-idle connections.  These are the values we use, on a pair
of systems handling a combined 3-7GB of traffic, daily, and
confMAX_DAEMON_CHILDREN set to 40 on each box.

	define(`confTO_INITIAL', `30s')
	define(`confTO_CONNECT', `30s')
	define(`confTO_ICONNECT', `15s')
	define(`confTO_HELO', `1m')
	define(`confTO_MAIL', `1m')
	define(`confTO_RCPT', `1m')
	define(`confTO_DATAINIT', `1m')
	define(`confTO_DATABLOCK', `1m')
	define(`confTO_DATAFINAL', `5m')
	define(`confTO_RSET', `1m')
	define(`confTO_RESET', `1m')
	define(`confTO_QUIT', `1m')
	define(`confTO_MISC', `1m')
	define(`confTO_COMMAND', `1m')
	define(`confTO_IDENT', `0s')
	define(`confTO_FILEOPEN', `1m')
	define(`confTO_CONTROL', `1m')
	define(`confTO_QUEUERETURN', `2d')
	define(`confTO_QUEUERETURN_NORMAL', `2d')
	define(`confTO_QUEUERETURN_URGENT', `1d')
	define(`confTO_QUEUERETURN_NONURGENT', `2d')
	define(`confTO_QUEUEWARN', `4h')
	define(`confTO_HOSTSTATUS', `5m')