Such a system enables transactions to occur between a limited set of participants without the requirement that the entire world know about them.
Avoiding telling the entire world reduces the cost of transaction processing by many orders of magnitude (something on the order of 1,000,000 times cheaper).
reduces the cost of transaction processing
Unfortunately Bitcoin has often been oversold as being "free"; but if transactions are free, and there's a fixed number of coins to mine, who pays for this whole network in 30 years?
The motivation behind the design of LN is to minimize the number of transactions which need to be globally verified (which has a cost) while still adhering to the general model of Bitcoin (decentralized, pseudonymous, etc).
You're right though that with tx fees being on the order of ~0.05 USD, you don't really need LN yet. I've long assumed that fees would become significant, pushing out some use cases. That's already happened with e.g satoshidice.
So you say LN is to solve future problems. Because transactions are free and in 30 years nobody will do the mining for free. Well, thats why bitcoin has fees. And a mechanism to let the market decide on those fees. This has been planned from the beginning.
From what I can see, all is fine with bitcoin. And there is no need for the complexity LN brings.
Basically any sort of payment you would want to do at the machine level. If I'm uploading a file to you, I would want you to pay per MB because I'm not sure if you're going to leave after you have 50% of the file.
Why pays to seed torrents? Who pays to run NTP servers? Who pays for running game servers? Its is wrong to assume that something has to be profitable in order for it to exist.
Altruism certainly exists, but assuming that altruistic actors (rather than self-interested actors) will expend more resources than attackers is a very different security model than what Bitcoin is designed upon.
This is causing nodes to go offline. There are about 500 to 1000 rasberry pies operating as full nodes on the network and their 1 GB of ram is not enough to hold the backlog of tx.
Blockstream is proposing to raise the limit to 2MB 2 years from now. That would be a disaster for the network.