we're having this issue since starting using CA RA, it was said to us, that this is a known issue. so we didn't investigate further.
what I experienced, but might also be coincident: when I'm not trying to update all agents at once via ASAP, but only select up to 5 a time, I didn't run into the problem that often.
we're currently looking into building an administrative service, that runs in parallel on all agents, which might be able to fix this issue automatically, when we trigger it