What, if any, errors do you see in the NW Java Logs on the MII server? My guess here is that either the ICM itself is saturated with HTTP requests or the MII Transaction engine (if you are calling is synchronously) is calling an action block that is throwing a timeout error, and this error is not caught by the "Catch" action so it is sent back to the calling system (ie: PCo).
Any insight you can provide on the NW Java Logs side is very important here to determine why the timeout is occurring. Which version of MII & NW is being used in this scenario?
Sam