Re-upping my post from 1.5 years ago on "layer 3s":

https://vitalik.eth.limo/general/2022/09/17/layer_3.html

The summary is that layer 3s don't magically improve throughput even more, though they can reduce some fixed costs of batch publishing and deposits/withdrawals.

However, there are other, potentially "lighter", ways to get the same cost savings that you get from L3s.

I hope it all stops at L3, even better at L2, and we won't have to read an article like 'Why build L4 when we can optimize/refine L2?' in the future.