If the block limit of 1mb will inevitably be eliminated, why refuse an increase in the size of the blocks at critical times like the ones we have seen these days?
Among other reasons: because, as our good Mayor implies in his reply, a hardfork is required to increase the block size. A hardfork is not unplanned for, but when it happens it should include a bundle of fixes, not just a quick bandaid.
Besides, the bandaid can't really solve the problem, even momentarily: if the spammers are willing to spam no matter what, they can as easily fill double-sized blocks with 1-satoshi transactions.
You are right, and the most efficient, in an ideal scenario, would be to make a hardfork with a series of improvements in the network; but the current conditions (spam attacks, new users, price increases, media attention, fomo) are having critical consequences in the network. Not only users have been harmed, but the stability and continuity of companies that have built their business model around the bitcoin blockchain (and at the same time add value to the network). For that reason, I think it is far from being a "temporary" solution, or of less importance.
On the other hand, it would be much more expensive for the attacker to fill the blocks.
Filling the blocks with transactions of ~ $ 20 in fees should mean a large sum of money (regardless of whether they are miners or not). Obviously, this type of attack is being made to manipulate the price. If the cost of carrying out the attack exceeds the benefits that can be achieved by manipulating the market, the attacker loses the incentive to carry them out. Also, if 2mb or 4mb blocks are filled with 1sat / B transactions, no one would care, because the fees would still be low and the network would remain decentralized. Jbreher recently mentioned the tx capacity of an average computer. I quote:
The only serious investigation into the matter has proven that your average 'home' computer today can handle a simple block size increase that will net us about 100 tx/s. And with a fix to core's crappy multithreading design, can handle block size increase up to about 500 tx/s. And that is without looking for other sw architecture improvements.
I would like to be able to express my ideas more widely, but the language barrier does not allow me to do so.