It appears “modular” means an incomplete design which winds up inefficient, due to security assumption miss-match caused by incompleteness, and also lacks a security definition for messaging.
We optimized bandwidth by tightly coupling availability to approval checking. If you “decouple availability” then availability should become more expensive, maybe in multiple senses, or less secure which requires analysis. If you’ve an optimistic roll up then you’ll pay some hefty latency price too.
If someone presents an optimistic roll up design which looks efficient and sane, then yes we’ll help them build it, but so far nothing.
We’ve plenty of more important much easier projects, like nobody doing real state channel work or faster storage. We expect 4 x improvement by fixing storage. We expect a dramatic improvement by using state channels, given state channels are bitcoin’s entire scaling story. An optimistic roll up has a similar unlocking delay down sides as state channels, but handles more diverse tx types and typically lacks a security model. We also know the stupid simple way of pre-committing to schnorr nonces, and limiting token flow, which also prevents double spending.