hmm we'd have to confirm that... but notwithstanding this attack
You don't need to confirm anything. Segwit makes validation time linear and prevents this attack (at current sizes).
there will be twice as many TX included pre block, requiring nodes to process relay and store twice as many TX, just as if we had simply increased block size.
Technically not twice, but yes that's right.
more is not always better?
It isn't. If it was, then we'd all run our nodes at the same locations and end up with a centralized network that could be easily shut down.