What does stateless Go mean?
A concrete and simple example: an api request comes in and you call n dependent services before responding. If any of those calls fail, the entire response 500s, if ok, respond 200. Instead, make this a job with a 202 response with an id returned and use a distributed queue and workers to async process the request and allow the job results to be queried by job id. Now any node failures are invisible to the client.