All the configs are shared across NFS. So config files are the same on
both OH1 and OH2...
the applications connect to mysql via php connect strings, and as tcpdump
reveals are doing the proper things
You might try using strace on an Apache process on OH2 and see if you can
spot anything
In your mysql setting increase the time-out period. With your application,
include a timing cycle that pings the server every certain time less than
the timeout period. That should remind MySQL server that you are still there
and alive.
Note this part of the MySQL Manual: