Apache process hangs in memory with mode operation L.

I have a heavy loaded website with 12req/second and I use apache 2.2.6 with mpm_event_module, because threaded and prefork modules suck already. Apache server with 5 mongrels in cluster: 3 on the same server and 2 remote. MYSQL is also remote. Memcached runs on the same server. NFSD also running. Below is apache tuning for mpm_event module:

StartServers 1
MaxClients 150
MinSpareThreads 15
MaxSpareThreads 25
ThreadsPerChild 25
MaxRequestsPerChild 0
It works great, with some exceptions: some apache processes hangs in memory with mode operation L (logging request to disk)
Mod status says that pid 31159 hangs for 1278 seconds (20 minutes) and it wastes system resources, making website to load slower. I had to kill hanging processes manually for some time, but in last weekend I wrote a simple ruby script that use hpricot gem to parse mod status and detect hanging processes wit h status L, then it kill them. I put it in a cron job to run every 10 minutes and i forget about the problems.