Should event_id always be a randomly generated UUID, or should it be somehow related to the underlying error/event/request in any way? I was planning to use the Trace-Id or Request-Id of a web-request as an event_id, but due to Are some event payloads ignored? Is this documented anywhere? I am not sure any more.
It should be wholly unique. We just use a uuid4 in all SDKs. I wouldn’t use anything like Request-Id because odds are it wouldn’t be unique if multiple events happened on the same request for example.