LN fundamentally changes nothing. Its still uses bitcoin Tx it just combines many Tx and does the on-chain settlement at a later point thus reducing the number of on-chain Tx increases speed and reduces fees. But if we reduce Tx on the chain we dont solve scalability we just bump into the limits a little later. LN on a global scale would still need orders or magnitude more on-chain Tx then possible.
And if LN would actually reduce on chain Tx to a level where we would no longer need to worry about on-chain limits then we rendered bitcoin useless because we literally reduced the need for Tx to basically zero. Whos gonna pay for mining if block rewards vanish and LN makes on-chain Tx vanish as well?
Bitcoins "goal" is that on-chain Tx should always be near the max so the fee per Tx is the lowest possible and reward are the highest While LNs "goal" would be to reduce the on-chain Tx to near zero because every on-chain Tx is extremely expensive.
It probably meets somewhere in the middle. which means its still gets worse and worse the more Tx are made even if they are made over LN. Thats the definition of "not scaling well".
LN and bitcoin are supposed to by a symbiosis but LN is the parasite that works best when it almost kills is own host.
You can see it though the actions of large miners. They are not supporting LN. They could pour millions in the development and advertisement of LN but they dont want it, it reduces their profit.
Its just very very deeply flawed overly complex system that only exist because some people would not accept that the layer 1 has deep flaws. Is rather obvious that building on top is not the right way forward. An since changing layer 1 is not possible in any meaningful way the only way forward is to abandon it. Which wont happen as long as people make money with it.
>what exactly would stop it from working?
The absolute end state is that hashrate drops or stops increasing due to the fact that it isn't profitable anymore which leads to double spend and essentially destroys any trust in the system and renders it useless.
No one know at which point hashing will no longer be profitable for a long enough time to cause this cascading effect. But once it started the price crashes > pushes more miners out of businesses > further reducing hash rate > panic > price drop more > repeat
All while mining hardware becomes cheap because demand goes to zero and supply increases, which makes a double spending attack cheaper.
Then once a double spend has been observed its game over. There inst even a theoretical plan how to move on at that point. In simple words it would move the requirement for block conformations to infinite so there is no point in doing any Tx anymore. Everyone just has to wait which further reduces any rewards for miners.