You subscribe to a list and/or document library.  A change to that list or document library has been made and an alert email notification is sent.  You receive the email message and discover that the email formatting looks like a combination of rich text and XML. 

The first thing I did was capture verbose logging.  What I found was nothing.  There's no error or clue to why the email notifications formatting is incorrect.  You aren't sure how to reproduce the symptom.  You're only left with theories, configuration issue, timer job issue etc. 

I decided to select a Preferred Timer job for the content database that was experiencing the issue.  In the environment I was troubleshooting there were 5 servers in the farm.  I picked one and generated an alert notification.  To my surprise the alert notifications were formatted correctly.  This lead me to believe that there was a configuration issue with one or more servers in the farm.  I performed the following steps:

1. Checked the 14/Resources folder on all servers in the farm looking for any kind of inconsistency.  (The farm I was working with had 5 servers in the farm)
I noticed that one of the servers in the farm did not contain as many resource files as the other 4 servers in the farm. 

2. In my lab (standalone environment).  I made a copy of the Resource folder. 

3. In the original Resources folder, I mimicked the Resources folder of the farm that I was troubleshooting.  The server with several missing .resx files. 

4. I then went to a list where I had subscribed to alert notifications, generated an alert and voila I was able to reproduce the above email. 

In the farm that I was troubleshooting, I recommended to copy the Resources folder from one of the servers with the correct amount of files to the server that was missing the files.  This particular customer that I was troubleshooting for was not ready to do that and opted to stop the timer service on the server in question that way when it came timer for this server to run the timer service, the issue would not suddenly reappear.  Later the customer ensured that all servers in the farm were configured correctly.  We never did determine how the servers got out of sync but my theory is during a CU update the patching was interrupted on this server.  The issue appeared around the time that a CU was applied.