Unfortunately, this would not fix the "bug", as I have explained multiple times. BRC-20 transactions, which are causing the congestion, aren't bigger than normal transactions with a handful of inputs/outputs (they have about 400-500 vByte plus another 110+ vByte for the commit transaction). They are an inefficient protocol with several WTF decisions (like storing data as text, and needing two transactions for a transfer), but not even that is the problem, the problem is their amount due to the "minting" hype.
BRC-20 transaction patterns could be identified regardless of their size, so a complete "block" is technically possible, Luke probably knows this, and wouldn't be risking dropping "normal" transactions for false positives. however, what Luke is trying to do now is add an opt-out feature to make your node reject Ordinals. Bitcoin allows all inscriptions by default, adding a new option to reject Ordinals-like transactions won't fix it, I mean even if the new core version comes with that opt-out enabled by default, people can still disable it, correct me if I am wrong.
Reading into Luke's comments here and there regarding this "ban" he does refer to it as "spam filtration" and he seems to be counting on the fact that most people including miners would activate the change, I find it very hard to believe that mining pools will opt-in for a new upgrade that makes them earn less.