Components:

1. ExecutionResult selection (https://github.com/onflow/flow-go/issues/7546)
2. Refactor lookup logic into modules: local vs execution node
3. Refactor execution node lookups to use passed in ExecutionResult instead of existing approach
4. Refactor local storage lookups to use new caching layer
5. New request parameters
6. New response parameters
7. Streaming fork recovery protocol
8. Implement improved Access API error handling

gRPC API Changes

The Access API will have the following new request parameters which are used to determine which ExecutionResults should be used to serve the request:

message ExecutionStateQuery {
	uint agreeing_executors = 1;
	repeated bytes required_executor_id = 2;
	bytes ancestor_result_id = 3;
	bool include_executor_metadata = 4;
}

• agreeing_executors specifies the minimum number of ExecutionReceipts for the ExecutionResult
• required_executor_id is a set of execution node IDs, one of which must have produced the ExecutionResult
• include_executor_metadata specifies whether or not to include the executor metadata in the response.
• ancestor_result_id (for future use) specifies an ExecutionResult that must be an ancestor of the given request. Used to ensure the user gets a consistent view of the state (or an error if the fork is abandoned)

Each execution state related endpoint will also optionally return the following in the response