For reference on the use case, see http://en.wikipedia.org/w/index.php?title=Special:Log&page=The+Dark+Knight+(film)&hide_review_log=0
Imagine three editors, Anna, Belle, and Charlie. Anna is non-autoconfirmed while Belle and Charlie are reviewers.
- Anna makes an edit to a page configured for pending changes.
- Belle and Charlie both go to review the edit.
- Belle finishes first, using the "undo" feature, causing the page to be reverted and automatically accepted.
- Charlie performs an edit on the page to restore the article to the original version (effectively the same as a revert), and chooses the "review pending changes" to accept his own change.
- The software recognizes that Charlie's edit was identical to Belle's so it auto-resolves the conflict by not creating an edit.
When this happens, Anna's change gets marked as reviewed by Charlie. What was supposed to have happened was Charlie's change would be reviewed by Charlie, but his edit doesn't exist (it is instead replaced by the automatically accepted change by Belle).
I think a more appropriate action in this situation would be to have no review done (silently would be OK I think) because the review was already done on the correct revision (just automatically).
Naturally, the outcome is still the same in that the page is reverted and the newest revision is accepted, but the history gets a little strange by saying an edit was accepted and then reverted. (It doesn't say that anymore in the example I gave above because I went back and revoked the acceptance.) Preferably this could be avoided, but it's a very minor issue.