This is because they are doing what is being termed as SPV mining (not normal or should I say "proper" mining).
When you mine in this manner you risk ending up on a fork (as happened last year to the Chinese mining pools after a soft-fork).
So if you had huge blocks and incredibly good hardware (and bandwidth within your pool of miners) then you'd just play games with those that do SPV mining and make them all end up on forks (after a while they would simply give up trying).
I hope you are starting to understand that SPV mining is *not good enough* (and the more you scale things up the worse it gets).
A couple posts back I had to correct your flawed understanding of SPV mining. If there is something I have said that is factually incorrect please refute it with evidence and not opinion.
SPV mining is perfectly good at mitigating the scenario you described in which a malicious actor creates a poison block on order to "own" the blockchain and collect all the rewards. Now that you have the correct understanding of how SPV works I think it should be clear to you.
Your premise that this bad actor would play games with SPV miners, was covered by my second question:
How quickly do you think other miners will ignore your 'malicious' blocks?
You said they had to accept the blocks because they were valid. However in the situation where the bad actor is gaming SPV nodes, then it quickly becomes obvious that although the block may be valid, the miner is malicious. So whilst the block may be valid, it is in the miners own best economic interest to reject all blocks from the bad actor. It is the bad actor who is forked off the network.
The reality of whether SPV mining is good enough is complicated and involves very many variables. So neither you nor I can make subjective judgements about its overall efficacy.
I would invite you to consider what happens with transaction fees in a 64MB full block scenario, and whether the risk/reward for SPV mining still favours mining empty blocks?
In doing so consider also these fairly reasonable assumptions:
64MB is a long way off even by the most optimistic big block proposal (e.g. in BIP101 we wouldn't hit it for another decade) work has already been done reducing the bandwidth necessary for block propagation, and thus the time. For those 64MB blocks to be full, then it would have to be viable for miners to be filling them. The size of blocks is self limiting according to the risk reward of including transactions vs being orphaned.
In 2026 the block reward is predicted to be 3.125BTC
If a 64MB block could be mined in a reasonable timeframe such that the chance of orphan was low enough to be acceptable to miners.
If the fee density remained consistent at 10k satoshis/k then the total fees would be 6.4BTC
FWIW, I think the incentive to mine an empty block is somewhat reduced as a result of scaling.
