It’s not a big surprise that microcomputers that were barely able of keeping up five years ago are pushing their limit.

it's a big surprise that the team of brilliant software engineers dedicating (a subset of) their careers to protecting the worlds most important financial technology from idiocy were dumb enough to let these perfectly functional microcomputers get swamped by memory constraints when they were working perfectly just a handful of years ago.

note that it's not the storage getting swamped, it's the ram.

This is an often repeated claim. Could you please provide any evidence that actually supports this claim? Obviously the UTXO set has gotten bigger, and that will cause Bitcoin Core nodes to flush the dbcache more often. On the other hand, this appears to be offset by spam transactions having significantly fewer signature operations per weight. BitMEX Research found a positive correlation between inscriptions and verification speed.

There are many things that people can do wrong that would make IBD unbearably slow. For example running your box with the default 0.5 GB dbcache when you have more RAM, setting the dbcache higher than 75% of your RAM, using an external hard drive, a SD card or HDD hard drive to store the UTXO set, running an outdated version that is by default configured to cease assumevalid at the height of that release and performs full signature validation for years worth of blocks. There is a lot of really bad instructions out there. E.g., I saw a semi-prominent RDTS proponent once suggest that one should set the dbcache larger than the UTXO set size (which if you have less RAM than that would cause swapping and freeze up your computer).
So, if someone at some point would please provide any evidence that the effect you’re claiming to see a) exists, and b) is caused by the spam, instead of just the blockchain being bigger because there are 100k more blocks than two years ago?
As long as nobody actually substantiates this claim, I’m gonna continue to assume that it’s a rumor based on some people misconfiguring their nodes and experiencing issues.

Before segwit, people were expecting blocks to land at 1.7–2.3 MB with full segwit adoption. The average blocksize has been about 1.6 MB in the last year.

it's the footprint of the UTXO set, not the size of blocks, not the drive size.

Larger UTXO set will translate to a graceful degradation, not a drastic slowdown. If you want to convince me of something else, please bring something more than repeating an unsubstantiated claim.

fee rates facilitating the creation UTXOs with huge witness is the problem. you don't want to un-discount the witness, and you don't want to solve for contiguous data. what would you like to solve for?

Please explain in detail which problem is being caused by huge witnesses.

Additionally, it is a fact that inscriptions and OP_RETURN outputs are cheaper to validate than payment transactions, and most of the additional UTXOs created are spendable no different from any other payment output.

why mention validation?

People claim (without evidence) that the IBD is gotten significantly slower since the spam fad started. Validation is a big part of what takes time in IBD.

their spendability is exactly the issue. UTXOs in the memory are what's constraining the decentralization onto weaker machines. the dominant cost on constrained hardware is UTXO set, not validation.

Could you please explain how the UTXO set supposedly impacts nodes? Because the points you’re raising makes me think that we disagree how it does.

put another way, which sufficiently powerful computer would you recommend for some poor user outside of the 1st world to save up for in order to run a node for their community with the preferred permissionless blockchain, hmm? certainly not a fairly common and inexpensive raspi4.

I would recommend that they instead get a used laptop. New Raspberry Pis are not particularly cheap for their very limited capabilities. After buying the device itself plus a harddrive and a case, you could get a pretty decent used laptop instead.

Spammers are only buying blockspace nobody else is demanding at minimal feerates.
... doesn’t give you the right to tell other people what they are allowed to do with the blockspace they buy.

they're "buying block space" at witness discount rates.

Blockspace is limited to 4,000,000 weight units. One byte of non-witness data takes as much blockspace as four bytes of witness data, so they also cost the same amount of fees. More witness data does not translate to displacing more other transactions. Witness data does not get blockspace at a discount, it gets a discount on its size. They increase the size of the block for the same cost. As explained, the blocksize is smaller than it was projected anyway and disk space is dirt cheap. Only the UTXO set has to be fairly quickly accessible, but that’s only about ~12 GB or so. So put those 12 GB on an SSD and put the blockchain on an HDD, if you’re running out of space?

tell me again why the witness discount was necessary?

The witness discount permitted segwit to increase the blocksize and it better aligned the costs of inputs and outputs with the burden they put on nodes.

(for the record, I don't agree that they're buying block space. miners sell placement priority in their templates not block-space... they can't sell something they don't own)

Miners produce blockspace and they sell the blockspace by matching it to bids from the mempool. The mempool is a perpetual auction for blockspace. Not sure why that would even be controversial.