This one's a doozy.

Steps to reproduce:

1. Create a revision with a blocking reviewer, and a non-blocking reviewer.
2. Have the non-blocking reviewer request changes.
3. Have the non-blocking reviewer then accept the same revision.

Expected result:

The revision's state is "Needs Review".

Actual result:

The revision's state is "Needs Revision".

Notes:

• The problem doesn't appear if there's nobody else that needs to review the diff before it can be accepted.
• I haven't tested whether having one person reject the revision, then have another person reject-then-accept, triggers the same problem, or if it is definitely a problem that is specific to blocking reviewers only.