Three reverts total -or- what is wrong with 3RR

What's wrong with 3RR

Essentially, 3RR is an electric fence - it automatically "zaps" any editor who crosses the line. Barring exceptional circumstances (reverting vandalism, enforcing WP:BLP, removing copyvios etc.), each editor is restricted to making three reverts per article per day.

There are several problems to this approach:-

Each editor's reverts are counted separately. This may be interpreted by a bad-faith contributor, someone who is willing to game the system, as an invitation to use sockpuppets or meatpuppets to win an edit war.
Each editor's revert count is reset each day. Although editors may be blocked for edit-warring, nothing in this rule prevents continued reverting day after day. The fact that another policy is necessary to address this issue should be an indicator that this policy isn't working well.
Each editor is restricted to three reverts per article, regardless of whether they are working on several problems or focussing all their attention on a single issue.
In limit poker, players are typically restricted to four bets and raises. For example, if player A bets $10, B raises to $20, and A re-raises to $30, then B may "cap the bet" at $40, with no further bets by A being allowed - although each player might secretly be hoping the other would back down and fold the hand. In case the similarity with Wikipedia's three-revert rule is not obvious: Wikipedia policy should not encourage editors to raise the stakes in this way.
Because of how reverts are counted, 3RR is fundamentally incompatible with the ideology behind the bold-revert-discuss cycle and the consensus model of article editing. Consider the following sequence of edits:-
Editor A makes edit.

Editor B reverts edit (1st revert for B).

Editor A reverts edit (1st revert for A).

Editor B reverts edit (2nd revert for B).

Editor A reverts edit (2nd revert for A).

Editor B reverts edit (3rd revert for B).

Editor A reverts edit (3rd revert for A).
From the perspective of WP:3RR, editor B must not revert again. In other words, editor A has won the edit war.

Some editors' interpretation of WP:BRD is that editor A ought to either begin discussion after B's first revert or accept that their edit doesn't have consensus. Others are of the view that editor B ought to begin discussion after making a revert. Both schools of thought agree that discussion should occur after the first revert and before the second.

From the perspective of WP:CON, editor B is reverting to a prior version, which is assumed to have consensus support. BRD provides for a discussion before the article changes from its consensus version, whereas 3RR prevents an editor reverting back to the consensus version.
A common extension of 3RR to areas of protracted or tendentious areas of dispute is WP:1RR. This restricts each editor to not reverting a revert, recognising that the problem isn't reverts per se but what is being reverted. By contrast to 3RR, 1RR does not encourage bold edits followed by repeated reverts that could be described as "bluffing" or "raising the stakes".

Of course, this reasoning may be criticised as only considering events from the position of editors who are contributing in bad faith, who are happy to game the rules, and who will not consider discussing the matter and establishing consensus. But these are precisely the editors whose conduct this rule must address. Reasonable good-faith contributors will (more or less by definition) stop short of engaging in an unproductive edit war.