The patch introducing experimental tumble did two things: removed the end delay on tumble, and added experimental tumble. The first thing is what made tumble way faster than running, the second is what makes it easy to do. If they revert both changes, tumble becomes useless again and nobody uses it again. If they revert just the second, they don't solve the speed issue. If the revert just the first thing, then the end delay makes experimental tumble painful to use, and probably worse than current implementation (because you can't double-tumble out of aoes, etc.).