To the best of my knowledge, SIGUSR2 is used to resume a suspended thread. Therefore, I would suspect that what's happening here is that due to the operating system having to handle tens of thousands of open file descriptors, a race condition is occurring where MySQL is asking for an available file descriptor from the operating system and is having to wait until one is available? Again, this is because of the *very* high number of tables needed in your server setup. That said, however, I am no expert on this type of stuff, and if one of the MySQL developers could weigh in here, that would be helpful. My best advice, other than getting an expert dev's opinion on this, would be to consider moving to a multiple db-server setup which can more effectively handle hundreds of thousands of file descriptors.
Let us know how things work out, and I'll keep my fingers crossed that one of the developer's is watching this thread. If you get no replies, perhaps consider posting this discussion (or the meat of it) to the internals mailing list and getting some opinions there.
cheers,
Jay Pipes
Community Relations Manager, North America, MySQL Inc.
Got Cluster?
http://www.mysql.com/cluster
Personal:
http://jpipes.com