The current active runner has been improved to parallelize the parsing of tests files, the idea would be to translate this to the new runner (might be a bit different as it is a different implementation but the current code could serve as a guide).