server goes pathological

Ubuntu-10.04, x86, stock.

I'm running into a situation where a server goes pathological. Basically, it becomes excruciatingly slow. I can get things through the file system, but a simple file touch can take 6 - 12 hours.

Top shows various btrfs processes hard at work - typically 100% of a cpu, (four cpu server), and the machine shows a load of just over 4. Some flush-btrfs, some btrfs-transacti, occasional others.

File system is was created with "mkfs.btrfs -m single -d single".

The same, (or a substantially similar), problem has arisen once so far on each of three different servers, so I have a good indication that it's not a hardware error.

Aside from rebooting, and rebuilding the file system, is there anything I can do to clear this state?

Is there any other information I can provide to help debug either what's going on with my server and/or btrfs?

--rich