Who is siding with spam?   Intended policy by whom? The very people who advocated for this particular limit initially  -- e.g. myself! -- support removing it now.  I was harassed for years over the op_return limit and even subjected to threats. Where the fuck were you then?

Aside, when I say spam here I'm just adopting the language of this discussion. I don't actually think it's a good description.  Spam is a message sent to you that you didn't ask for and almost certainly don't want by a second party who hopes to profit from it and who paid essentially nothing to send it causing you to waste a lot of resources reading it. You win against the spammer if you don't have to read the message even if your computer processed it or other people read it.

This normal definition spam doesn't exist in Bitcoin except perhaps for dusting which almost no one cares much about.     The stuff that people are calling spam in Bitcoin is where two consenting parties transact with each other entirely consensually, and they pay a third party handsomely to process it.  This irritates some Bitcoin users because it consumes network capacity (like any other txn) and does so for the benefit of some activity the user deems to be not sufficiently Bitcoin related. To defeat it it isn't enough that the user personally don't see it (they wouldn't have anyways), but rather they must assure no one sees it/processes it because Bitcoin is a consensus system so as soon as one miner accepts a valid transaction all participants must accept it.


In what way does removing this limit harm small miners or encourage centralization?  What are the specific 'down the line' dangers you're referring to? Surely they can be enumerated if they're significant.

It's easy to enumerate the opposite:

With the limit in place, miners that bypass the limit will earn more income than others. They get the transactions via direct submission.  There is little reason for someone to direct submit to small miners, since the large ones do the job and the small ones may well be (and hopefully are!) anonymous.   Lots of software authors already somewhat prefer direct submission because submitting to an API fits better with JS-jockie programming practices than submitting to a P2P network,  if the latter is also more reliable at getting their tx mined then convincing them to do otherwise is a losing battle.   In any event the consequence of direct submission is that the biggest few miners make more income, an because mining is highly competitive by design this ultimately tents to push smaller miners into operating at a loss.

Similarly,  when txn that will get mined don't get relayed in the public p2p network then miners that learn of blocks via the network (e.g. especially small and/or anonymous ones) then they propagate slowly.  Slow propagation causes mining to be more race like and less lottery like.  In a lottery you win proportional to your tickets (hashrate).  In a race the fastest always or almost always wins.  If blocks always transferred instantly then mining would work as it is ideally imagined like a perfectly fair lottery.  Slowdowns increase stales for everyone but cause smaller miners to experience more stale blocks than larger miners, and as difficult adjusts larger miners make proportionally more than smaller miners. Another centralization pressure.

I don't think these points can be denied though it's not unreasonable to debate their significance.  However, on the other side what is the downside to weigh them against?   The limit is already functionally not enforced. The spammers already have alternatives that are better for them and/or worse for the resources needed to run the network.