At the very least there's no guarantee of that. If you send a conflicting transaction with a higher fee, a miner should (rationally) discard the older one. And last I checked (Ethereum, a few years ago), that was indeed what they were doing.
If miners did change their Bitcoin nodes to facilitate double spending even with RBF disabled, that would erode trust in Bitcoin and thereby de-valuate it, which is probably not in the interest of miners. They'd damage the thing they're heavily invested in.
Description of the "replace by fee" mechanism you're mentioning: https://en.bitcoin.it/wiki/Replace_by_fee
This page mentions that it is a "node policy", and from your comment I think we both agree that nothing in the protocol (which roughly guarantees that Bitcoin is fine as long as 51% of the hashing power belongs to benevolent actors) ensures these rules are in place.
Now, you're saying there would be some reputational damage which leads to devaluation which is bad for Bitcoin, so bad for miners, so they wouldn't do that.
This will not happen until such payments represent a big share of Bitcoin transactions. In particular, if a few merchants were to start accepting transactions that are not mined and just propagated, they could easily be "victims" (again, it's economically rational for miners) of double spending and there would be no impact on the valuation of Bitcoin.
Nobody forces merchants to stick to requiring block confirmations for e.g. Bitcoin transactions.
They just do it because that's the way it always has been.
Instead of that, for low price purchases, they could also monitor the network for how well-spread a transaction is and once it is sufficiently propagated assume that a double-spend won't be possible anymore because miners will refuse to accept a transaction which conflicts with the one they saw much earlier.
And "well-spread" would be typically the case within seconds because transactions are flooded to all network nodes as they appear.