So, we've had this problem forever on our servers. We do not have Symantec support anymore. But when we did, I use to call them every month or two with this problem where a server would go from having a handful of MAPI profiles to having thousands overnight which breaks everything. I usually go into the registry and delete the Profiles key and restart the services. But they had a fix that I can't remember what we did that made the problem signficantly better for a while. We had the problem very often, EV support instructed us to make some change and the problem happened signficiantly less due to the change. I think it was a registry change. Lately the problem has been happening a lot more.
Some reason, the last few times the problem occurred, deleting the Profiles key didn't work. I'd reboot the server and the key with all the thousands of sub keys would return as if I never deleted it. A few errors ago, I discovered that our service account appears to have no significance to the server running. I'm sure the permissions mean something assigned to the domain service account, but there's nothing local in the account that has any bearing on the server running correctly.
After not being able to delete the profiles key with regedit or powershell and having the server lockup for hours trying to delete it, one of the time, I rebooted the server and a temp profile loaded. I noticed that hkey_users was a default profile and everything was still working fine. That clued me that the real hkey_users assigned to our service account meant nothing. I then disabled the ev services, rebooted the server, logged in as a different admin account and deleted the service account for enterprise vault, restarted, started the services back up and all the corrupt mapi profiles were gone (because they were in the corrupt hkey_users registry.)
It's been fixing the problem. But the problem is reoccurring more often lately. I'm wondering if the original fix we did years ago was in Hkey_users so me deleting it is getting rid of the fix that was causing it to happen less often.
BTW, we're running EV version 9.0.4.
You could try implementing the following in addition to what Virgil and Gertjan stated.
How to restart Enterprise Vault MAPI Tasks automatically.
For example setting RestartAllMAPITaskIntervalMins to 10 hours or less helped a lot in some environments.