More of an FYI in case anyone has run into this problem...
We were running into the infamous "timeout on buffer latch type 3" error and
was caused by the autogrow in tempdb. Apparently when SQL Server is getting
hit pretty hard and tempdb is taking a timeout to auto-grow, SQL server
craps its pants and times-out. This combined with the fact that tempdb
shrinks down to its last set size after a restart, you'll hit this error
pretty regularly after a SQL Server bounce. (If tempdb auto-grow from a set
1GB to 10GB, after the restart tempdb will be 1GB again. - Just how gay is
that?) Anyway, we just manually set tempdb to a respectable level to resolve
the problem.
