@MattBurrows In late January 2020, Identity Protection and MCAS started integrating, so my guess is the change in behavior you observed on 2/26/20 may have resulted from the code change from January as described here: https://docs.microsoft.com/en-us/azure/active-directory/fundamentals/whats-new#two-new-identity-prot...
There is not perfect parity between MCAS and Azure Identity Protection. For example, I asked Sarah Handler (MSFT):
Q: "Azure Identity Protection already had impossible travel signal before MCAS sent its signal so are you eliminating redundancy and using MCAS’s engine for that now? If not, How does this seemingly redundant impossible travel signal add any extra value to AIP? Was AIP deficient?"
A: "The signals have very similar names, but detection logic is distinct and they are looking at different things to determine compromise. There may be overlap, but we’ve definitely seen cases where AADIP will see atypical travel and MCAS doesn’t see impossible travel and vice versa. Atypical travel is the AADIP signal and Impossible Travel is the MCAS signals. 2 things have changed in the last year: 1) we previously called the AADIP signal “Impossible travel to atypical locations” and renamed it to “Atypical travel." We integrated with MCAS to consume their Impossible Travel signal, which now shows in our UI and can influence user risk"
https://twitter.com/ITguySoCal/status/1232862860084072448
The Unfamiliar sign-in properties risk type is continuously updated.
There was a big update in August 2019:
https://techcommunity.microsoft.com/t5/azure-active-directory-identity/presenting-the-new-unfamiliar...
And then another update in November 2019:
https://techcommunity.microsoft.com/t5/azure-active-directory-identity/the-refreshed-azure-ad-identi...
