We have a problem in our DSMA build that the DMSA service keeps experiencing a CPU leak - the CPU ramps up to a set percentage (ususally 33, 66 or 99%) and then stays there, probably as a result of many interfaces all launching at the same time. We have moved the interfaces onto a dedicated DSMA service Provider of their own (we called it DSMA Automate) and it has two advantages.
1, it appears to have reduced issues with the Interfaces getting locked up .
2, If the DSMA service does experience a CPU leak, we can kill the service and restart it without having to stop the users from working. This means we can solve the issues in a live system without notifying the client of the issue and arranging an outage window.
This solution is working reasonable well, so I think there may be an advantage in building it as a default build.
Also see my service ticket number 11021