I was curious so I tried it, and it filtered straight through on my synchronous AG. Cool. (I changed from 140 to 130 and back again). No need to remove the database from the AG.

The change was reflected in the error log of the primary node but not the secondary (which makes sense).