Do what you are doing but dump the contents of tracing into an LLM agent (cowork, code, opencode, etc) and ask for it to take a first pass. It’ll at least narrow it down for you. Use a smart model and it should be helpful.
Yeah, Claude's cost really adds up fast on multi-step traces. I haven't tried OpenCode yet, but I'll definitely give it a spin to save some API credits. Thanks!