>the miners must select one chain to mine
I am pointing out that the "sloshing" is in fact perfect evidence that they are only mining one chain at a time, or splitting their hashpower to varying proportions. This is completely normal and healthy miner behavior, they are simply greedy actors looking for the most profit. They cannot however mine both chains with their full hashpower for free (as you can with PoS), and this is an essential understanding when analyzing the security model of a consensus algorithm.
Also, the oscillations are most commonly exploitable only when the difficulty adjustments happen very quickly (see: BCH's EDA), so if 70% of the hashpower left, even if 16% of the remaining hashpower was malicious, the efficacy of a 51% will be handicapped by the disproportionately high difficulty.
I don’t care whether sloshing is “normal healthy behavior”. I care whether a fork has become easy to attack. You seem to focus on every topic but that one.
Original comment: "Also, with Proof of [Work] you get protection against an explosion of too many hard forks..."
My reply: "That's not what happened with the Bitcoin Cash fork though ... any given time, one of the networks is easy[er than 51%] to attack."
If you're not disputing that point, then you're not refuting my objection to that original comment.