What’s the future of L2s if Ethereum L1 gets near-zero fees?

With Ethereum’s roadmap (blobs, DA scaling, statelessness, etc.), it feels increasingly plausible that L1 fees could become low enough that most users won’t feel pain anymore.

That raises a genuine question I’ve been thinking about:

If L1 fees become negligible, what role do L2s play long term?

More specifically, do we expect:

High-speed generalized L2s (MegaETH-style: ultra-fast, general purpose, pushing the execution frontier) or

Hyper-optimized L2s / app-focused rollups (e.g. Lighter and RISE – trading-first, CLOB-native, synchronous infra, custom execution environments)

…to dominate in that world?

Some thoughts I keep coming back to:

If cost is no longer the bottleneck, latency, composability guarantees, and execution determinism might matter more than raw throughput.

Certain apps (perps, on-chain orderbooks, games, HFT-like strategies) seem fundamentally incompatible with L1 block times, even if fees are cheap.

On the flip side, generalized L2s risk recreating “mini-L1s” unless they offer something structurally different beyond speed.

So I’m curious how others see it:

Do L2s remain primarily a scaling layer, or become specialized execution layers?

Does Ethereum end up as the settlement + coordination layer, with execution fracturing by use-case?

Or does cheap L1 eventually compress most activity back to mainnet?

Would love to hear perspectives from builders and researchers here, especially how you’re thinking about this post-cheap-fees Ethereum world.