For web-style application code my experience writing Go code says that you totally can stop using logs and just use traces and have just as much visibility into a running process, if not more.

For library code or other deployment models I’m not sure - it’s not something I’ve had a need to experiment with