wait for a response.  Also, use a single cache list command to check
the MD5 of both files.

This speeds up startup immensely.  It would be a good idea to do
something similar in Synchronet any time we send a query and wait
for a response... as it sits now, we're basically adding extra
delays that aren't strictly necessary.

Another option would be to *always* flush before a blocking read,
which would "fix" anything that's not using inkey() with a timeout
of zero.  This would just generally fix everything I think, unless
there's something calling inkey() with a very small timeout just
to avoid 100% CPU or something.