[
Let's consider the most efficient option: one LN node and many end users connected to this node. The node must invest the same amount as all users do. If the node wants to earn 3% of the profit for the year, then all users will have to pay this 3%. That is, on average, each user will pay 3% per year of the amount invested in the channel. (The average route length will be 2).
3% sounds bad if you spend $1000 per month. But if you put $100 in a channel, and spend and receive $10 at a time, but do that 50 times per month, then $3 in fees doesn't sound bad at all! I spent more on my last on-chain Bitcoin transaction (I made a mistake on the fees, my fault).
Again, an unrealistic scenario.

I (like most users) don't get paid$ 10 many times over the course of a month. I usually pay$10. After 10 payments, my channel will be reset to zero. And you will need to either re-create the channel, or replenish it using a third-party service. And this is the fee for an additional on-chain transaction.
"Super-cheap transactions in LN" and "LN-nodes that make a profit" are mutually exclusive things.

You're probably right here. But I don't really mind, it's not as if Bitcoin nodes are currently making a profit. Some things are just done to support the network, and if an exchange (or casino) runs a LN-node, earning money from transaction routing is only a side-income, and not their core business.
Is it unprofitable to keep an LN node? This means that there will be few nodes, and their number will not grow. What we have now.