how well do the agents handle complex multi-service bugs in large codebases? | discoverkit | discoverkit