We experienced the 'lost input channel' on rcpt messages after we migrated
from Listserv 1.8d on AIX to Listserv 1.8e on Linux-S390.  We did not get
any (or maybe I missed them) 'resource temporarily unavailable' messages.

All indications were of a shortage of a resource but we couldn't figure out
what.  Since it was on an OS platform that was completely new to us, we
thought it had to do with performance issues on the OS rather than a
listserv issue.   We kept tweaking memory, disk, etc. without success.
Analysis of the sendmail logs showed that the lost input channel
consistently occurred on large lists so we lowered the maxbsmtp setting and
got some relief but not a complete resolution.

Further investigation revealed the lost input channel tended to occur after
X number of rcpts.  A manual dns lookup of Xth recipient host address
always gave an error.  So it was a DNS related problem.  Still not fully
understanding the underlying nature of the problem, we were able to
circumvent it by setting FEATURE(nocanonify).  We assume that this improved
the handshaking between Listserv and Sendmail so that response was faster
and 'something' didn't timeout.

Good luck!

Janet Shih
Louisiana State University