Mempool policy makes it inconvenient for people to use transactions that
violate the mempool policy. It may discourage them from building
protocols that require such transactions. But this discouragement has no
monetary value, which means that as soon as there is any economic
interest in producing such transactions, they will be produced and they
will wind up in blocks. This is what we see -- and it's why we are
talking about eliminating the data carrier filters and not about
eliminating, say, the MINIMALIF rule on pre-segwit transactions.
I fully agree that this discouragement has little monetary value. As we can see today, folks are bypassing the existing default mempool minfree rate of 1sat/vbyte and still managing to fill up blocks with these sorts of sub-1sat/vbyte transactions. This lets us measure the monetary cost of bypassing default mempool filters. The cost is less than the cost savings that such folks are achieving by their sub-1sat/vbyte transaction. Eyeballing it, I see that the cost of bypassing default filters is something less than 0.3sat/vbyte or so. Probably there was initially some upfront cost, which is now being amortized.
I'm glad to hear that the default minfree rate is being lowered. I'd even support eliminating the minfee entirely and instead relying on the max mempool size alone.
Also, as Andrew touched on, one valuable role of filters is to try and filter out third-party malleable transactions to the extent reasonably possible, or at least filter out their non-canonical / non-min-cost forms. That is valuable because those sorts of transactions are at great risk of never appearing in blocks.