[Brian]

We got a couple of sample databases from folks, and we think we may know what's going on here. Can folks that were/are hitting this problem let us know if they'd used the new 'delete old client files' feature we added recently?

From what we're seeing, it looks like we're spending all our processor time compacting the database. Our theory is that the new feature means that some folks with really stale client files deleted them - consequently, we have a bunch of large databases (since they haven't been compacted for weeks or more) that can suddenly compact.

If that's correct, folks never hit this problem before because they never compacted - they just added new transactions to the pile with every sync. Compact fairly regularly, and things don't appear slow because you're only compacting a small number of transactions. Throw a thousand or more transactions in several branches that weave in and out of each other at the algorithm, though, and things get slow.

The lead developer is going to work on making that more efficient tonight - when I know more, I'll make another post.