EL parity test suite of rollup-boost. #423
Description
Based on our recent experiences with facing batcher issues just because of the fact rollup-boost broke parity with how certain errors (request timeouts) are responded by op-geth, it would be a good strategy to have some sort of parity/regression-testing suite which would validate the consistency b/w what rollup-boosts responds v/s what the actual EL responds.
What we faced?
- The 0.7.6 version of rollup-boost started wrapping "request time out" errors as RPCError objects, unlike before.
- Batcher responds differently to errors when they're of the type RPCErr.
- And at the rare times of unavailability, although geth responds to "request timeouts" as a non-RPCErr, rollup-boost relayed them as RPCErr to batcher (when serving miner_setMaxDaSize RPC) causing it to behave in a fatal manner, which caused downtime.