The GRPC message limit is designed to prevent a client consuming unexpected memory by being pushed a large response by the server.
Given this intention - it is applied on the uncompressed message, so compression settings do not save us here.
I havent been able to measure directly, but I believe GRPC does deflate compression by default and is likely compressing these messages over the wire very effectivly.
Previously we have worked around this limit by creating streaming versions of the endpoints, but then we just concat all the slices together anyway so why not just raise the limit and receive the whole payload directly.