The problem seems to be reproducible. After finishing 16 % of the
dump, it jammed again.
Hmm, after unloading and reloading the medium, it works again. The
degenerated performance and start-stop behaviour are gone now.
Weird. Smells like a bug somewhere, but /where/?
Could it be related to the SCSI bus reset that was needed to revive my
pt0 device (see other thread)? The tape drive is on the same bus. Of
course, the tape was idle while i issued the camcontrol reset command,
but the cartridge had not been unloaded afterwards (until right now),
so for sure sa(4) must have seen a unit attention when going to access
the tape again.