Of course, we are getting into very large levels of speculation, but just on the economic level, you are really exaggerating and even seeming to lack imagination.
Eh, you can call me a bit of a pessimist. 
Let's just say hypothetically that on chain transaction fees were consistently $1k, so of course, it would not seem correct to suggest that you need $1 million to create a lightning channel.
That's not what I meant. My point was that if it takes $1k to make just one (~150 vb) transaction, then you wouldn't pay that fee, unless you owned 1000x that amount. We surely can't expect from the middle class to open lightning channels in that case. maybe we are also a long way from getting to some kind of a status of persistent $1k onchain fees, too?
Well... If block size limit remains at 4 MW, then to retain at least half of the current reward, transactions should be paying about 100 sat/vb each on average. This would ensure roughly 4 BTC block reward. Both an opening and closing transaction weight about 330 vb in total. That comes at 33,000 sat per lightning channel on average. That amount might seem low in terms of the current exchange rate (~13$), but it can be pretty horrifying if we ever exceed 5 or even 6 digits. Define micro-transaction!
Judging by the previous discussions of ours, you must have comprehended my spirit on that topic already. My opinion on what is a micro-transaction is irrelevant. But, if you really care about my opinion, here's how I see it: A micro-transaction is any transaction which can happen on a second layer and come less expensive than on-chain. - A transaction that transfers 1000 sat is cheaper in the lightning network.
- A transaction that transfers 1 BTC is cheaper on-chain.
Lots of things can influence the final fee (many inputs on-chain, high charges off-chain etc), but I'd say that about $50 to $100 is what draws the line between off and on-chain currently.
meaning spammers are penalised for spending too often.. rather than the current model where everyone is penalised because spammers fill blocks
Whereas in your model everyone will be punished for spending too often! Great! 
This changes the entire model, and as already said requires a softfork at least. Needless to say that we shouldn't punish people who spend money at times disapproved by franky1, the sole arbitrator of rightness.