28314

If you pack a thousand simulations into one request, how should resource limits be calculated? Reth PR #28314 proposes shared constraints for three types of callMany interfaces: at most 1,024 calls, one rpc.gascap budget shared across the entire request, and no new calls started after five seconds. Exceeding the total gas budget causes the request to fail, and operators can adjust the configuration. My view is that batching is just a way to organize requests; clients still need to control the total cost of the batch. $ETH When designing batched simulations, applications should split them according to actual node limits and preserve failure information; “no new calls started after five seconds” also shouldn’t be interpreted as a guarantee that every call will finish within five seconds.