Comment 23 for bug 118625

The "first slowdown" from the above comment #20 is a place where loggerhead attempts to build the revision cache four times simultaneously for the same branch (namely, a branch of launchpad itself). I wouldn't be surprised if you could reproduce this simply by opening four or five revisions in a tab for a branch that didn't have a revision graph, on codebrowse.

I can confirm this behavior as being constant every time there's a hang, looking at various logs that I was sent by spm: Simultaneously make several requests that will cause one large branch's revision graph to be built (mysql, launchpad itself, bzr itself) in several threads simultaneously, and loggerhead will slow to a crawl. It doesn't actually seem to *hang*, it just gets so slow that it seems like it's hung.