LN runs offchain, pretending that IOUs are the actual funds, they are merely IOUs.
As such, the IOU value is a pretense of LN code , and code can be modified/updated or tricked in some instances.
This is not how Lightning works.
If you and I open a Lightning Channel, each with 5 BTC in the channel, there will be a funding transaction on the blockchain totaling 10 BTC. Each time we transact, we will both sign a new closing transaction, that we do not broadcast, that would "settle" the channel with the amount of bitcoins each of us will have upon closing the channel. At no time will the closing transaction reflect either of us having more or less than our respective balance, and the total will never be more or less than 10 BTC (less transaction fees).
Anyone, including you or I can write any proposal we want, and can write any code we want, even if we know it would not be a good idea. That person is proposing in a nutshell a way to make loans via the LN.
2. The top 4 mining pools could 51% attack btc and double spend,
this would trick LN into loading channels with bitcoins, that would be redirected to another onchain address.
Yes this is a risk. There is this same risk if I were to buy bitcoins from someone in exchange for my cash; I could receive the bitcoins first, wait for a couple of confirmations, give you my cash, leave and realize the transaction is no longer valid after the transaction I received was double spent after a miner orphaned many blocks.
Can't one hypothetical day the amount of new members per day be too high that the setting funding transactions on the blockchain (soon or later) overloads the network anyway?
LN will not allow bitcoin to scale to infinity. The number of channels that can be opened or closed per day is not very high, but once channels are open, a very large number of transactions can take place over the LN network.
Additional technology improvements needs to be done before bitcoin can scale to Visa levels, but there is no need to create a solution that scales this much now because we are far from having that kind of demand. If I am in California, I have no reason to download every transaction someone does in New York, or in Western Europe. Having every transaction recorded on the blockchain is unnecessary.
If the cost to have a transaction confirmed on the blockchain is $2, if I wanted to buy a song on iTunes, or a cup of Coffee at Starbucks, there is no way I would use an on-chain transaction because the cost of a confirmation would be as much as what I am buying. If I already have a LN channel open, I can buy either of these items via a LN invoice that costs small fractions of a cent to pay, or if I do not have a LN channel open, I could pay $2 to open one, pay via a LN invoice, and pay for many other things over time, each costing a small fraction of a penny.