While testing one of my programs I found a significant difference between my results and the ones from the CLT. I thought I had a bug, but after a lot of debugging it seems that it is instead the CLT that is buggy.

Case in point: I searched the CLT for "Paull Walia" (note Paull with two L). One of the profiles listed by the CLT is the German release of Ocean's Thirteen, UPC 7-321982-000868. It shows Paull Walia as "Security Technician". However when I downloaded the profile to check it, I see that the name in the actual profile is really "Paul Walia" with a single L. 

Ok, I wonder if T!M updated more than one profile, because I got perhaps half a dozen mismatches.

This week, I've contributed to exactly 12 'Ocean's Thirteen' profiles that each had woefully incorrect sets of cast and crew data. I did that while working on the common name for Paull Walia, indeed - see my common name-finding thread for him here. Those 12 'Ocean's Thirteen' profiles each listed "Paul" when the actual credit is for "Paull", which made it incorrectly look like like Paul Walia was the common name, which it isn't.

The CLT isn't "broken" (at least, not in the way described here), but there's a difference in the "live" CLT data (which immediately changes the second a contribution is being accepted) and the updated database-files you're probably using for your programs - that data is only updated once a day. So at any given point during each and every day, there can always be differences between the daily-updated-data you're drawing from, and the "live" data the CLT is drawing from. It's always been that way.