Was this the one you replied to a bit ago about 'rogue' devices?

   With your advice in mind regarding possible cranky firmware on the cdrom,
   I disconnected the cdrom, and the esp probe timeouts stop.

CD-ROM drives are notorious for having bugs in their SCSI protocol
implementations, often relating to the use of LUNs.  If you send a
MESSAGE IDENTIFY with a LUN other than 0, some will immediately go to
BUS FREE phase (causing retries and an eventual timeout), and some
will simply pretend you said LUN 0 and even reselect from LUN 0 later
on.  And this from drives that report themselves as being SCSI 2.

The only way around this is to not probe at LUNs > 0 on those drives,
or to try to fail more gracefully (but this is almost impossible to
do).

   I'll be posting the errors as soon as I can get the machine back up.

In your case, this is easy to work around by putting an entry in the
`quirk' table in scsiconf.c.