Are L2 Rollups really the panacea for scaling?

I'm trying to get a better understanding of the roadmap, because I don't understand how Rollups are really going to work for real-world high volume usage.

Let me clarify that, real world, cheap, secure and convenient usage.

Each L2 is like a silo, in more accurate language it has no asynchronous composability with the L1 or other L2s.

If a dApp is on L2 A1 and your funds are on L2 B7, you have no access to the dApp. You can go to L1, and back to L2 A1, but are people really going to do that cost and inconvenience?

Other solutions I'm reading about are inter-rollup bridges or intermediaries. Even if these are fully secure and invisible to the user (will they really be?), then you still face two possible problems:

Routing

Cost

The issue being unless every L2 is direct bridged to every other L2 you might need multiple hops to get from Rollups B7 to A1. The routing problem is well known, with no centralised party managing route discovery, how is the optimal route discovered?

Cost is that every bridge/intermediary has to be created, liquidity maintained, and that ain't for free. Connecting all L2s to all others seems very inefficient and in itself creates another layer of cost/complexity. Similarly publishing every dApp on L1 and all L2s seems very expensive and fragments liquidity (I believe?).

A core innovation of CryptoCurrency is dis-intermediation; the removal of intermediaries. That's what makes it P2P.

Hopefully these issues are thought through and someone can link to a clearer definition of how Rollups scaling is expected to interoperate at mass scale.

TIA