This wasn't really resolved, and we're waiting on Killdozer to look at things again.
This is what our host writes about the
----------------------------------------
I'm really not sure what the cache directory is for, but it's the cause of a great deal of disk activity which is why I asked if it's safe to have a shared cache directory (a question you need to ask the TradePulse people). It does seem like TradePulse is meant to have one install per server, not per domain, which may be part of the problem here. Unfortunately we do not have much experience with the software and it's encoded which blocks most of our attempts to examine it.
----------------------------------------
The cache directory we're talking about is:
/tp/data/logs/cache
We have around 30-40 tradepulse installs on one server, and over 100 on another (not using thumb grabbing, just text toplists on the 100+ one)
Our hosts also writes this:
----------------------------------------
Besides the disk IO issues it seems that TradePulse is occasionally making HTTP requests to a third party server that is fairly slow.
While doing this, the script consumes 99% CPU and causes other processes to back up causing the overload condition. I've located an auto-update setting in TradePulse which I have disabled, I have not yet been able to determine if this was the cause of the spikes.
----------------------------------------
I'm not sure but I'm thinking this might be the thumb grabber?