Problem

When GW RPC clients send a large number of failed transactions it can have knock on effects and use up EVM Gateway keys at a faster rate than is preferred. This in turn can create operational overheads for Flow.

One way to mitigate this could be to support address based rate limits (only for failed transactions?) to mitigate the potential impact of a rogue script or overly aggressive automation that isn't being monitored by whoever is submitting the transactions.

Acceptance Criteria

• Caller to the GW RPC who trigger a high rate of transaction failures are rate limited
• [Could/should we communicate with them and ask them to reach out in the error message we send?]