Yeah, I’ve been crunching the numbers. Rollups really boost throughput by packing dozens of transactions into one L1 commitment, so you get a big drop in gas cost and a huge spike in TPS. But that’s only true if the data availability is solid—if the rollup can’t prove the state changes, the whole system stalls. The pattern I see is that optimistic rollups get more throughput because they only submit data, not proofs, every block. They’re cheaper but vulnerable to fraud‑proof delays and potential censorship from the operator. ZK rollups trade that for instant finality: the proofs are heavy, so they’re a bit slower per block and more expensive to generate, but they lock in the state quickly and eliminate the fraud‑proof window. So, throughput scales with compression and batch size, while security stays tied to L1 and the operator’s honesty. The sweet spot is somewhere between very large batches and a fast, robust fraud‑proof system. It’s a classic trade‑off: higher TPS, higher risk.

That’s the big lever right there. I’m still profiling a few new proof‑generation schemes; the idea is to shard the proof construction so the bottleneck splits. If we can cut proof time from a minute to, say, 10 seconds without adding a lot of gas, the whole window for fraud‑proofs shrinks to a hair’s breadth. I’ll ping you once I’ve got a fresh set of latency metrics and a rough cost curve. Stay tuned.

Got it, I’ll send you the raw timing and cost figures as soon as they’re ready. In the meantime, if you spot any anomalies or patterns in your own crunching, drop me a line—two brains are better than one when hunting for optimisations.