Unfortunately, in 3.0 and 3.1, changes to our UI impacted error handling throughout the system. We addressed and fixed this problem in 3.2 (CSCva10992).
503 errors, in most cases, are because too many requests come through in a given window of time or because too many requests are being processed concurrently. 503 errors should not be sticky (that is, they shouldn't persist forever). If you hit a scenario where 100% of responses are 503s, often times it is because the system clock has been rolled back, which messes with our accounting of requests.
To answer your question, on 3.0 and 3.1 we do still capture the exact cause of the 503 response in the xmpNbiFw.log file in the PI log folder (/opt/CSCOlumos/logs).