Ran into a strange one. All messages from outside were AWOL. Found them all sitting in commserver\unitymta. Found another thread that talked about bouncing AcUMRSyncSvr. Bounced it and all 600+ messages were delivered in a few seconds.
Exchange environment is Exchange 2003 Enterprise in a 4 node Active/Active/Active/Passive cluster. Looks like the partner server (EX03) had been failed over to another node around the time the problem presented. Actually, it had been failed over a few times during the night as the exchange crew were working on the cluster. Each time, the UMR service went offline and then reconnected when it came back online. The last time, it had a new error message with Event ID 144:
"Could not retrieve Outside Caller Object. The UMR will not function correctly until this is addressed. Please call technical support. 0x00000F2C"
Outside callers could leave messages. Inside users would not receive them, nor would they get MWI.
I think, but need to verify, that they didn't get any indication when dialing in to check their messages that a message was waiting or the standard message that comes up when the partner server is down. I was working it remotely, so I'm not positive about this.
In researching, I determined that the partner server is not running MTA. I found an article that mentioned to use the cluster node that is running the MTA. Is this really the only issue? I can re-home it, but I just wanted to make sure we didn't have a bigger issue that could cause this problem to crop up again.