Thanks for the clarification.
But would Segwit have activated without the initialization of BIP148, or would the miners have continued blocking it?
I believe "another UASF" would have come.
I think bitmain just wanted time to dump their covert-AB only hardware on a willing sucker (like, say, Calvin).
As of my understanding, AsicBoost is not defeated by segwit that much. I don't think on-the-fly transaction re-ordering is necessary at all for AsicBoost while it is exactly what Segwit makes
harder.
A hypothetical AsicBoost-only miner machine basically doesn't need and can't change the header on-the-fly as it is received from the pool operator in the most practical cases. I'd conclude that on-the-fly transaction re-ordering efficiency is just important for
solo miners with AB-only machines and even for such miners it is absolutely possible to keep their machines busy mining empty blocks in a few seconds window while their cpus are preparing a well-formed header with hundreds of transactions.