It's a good start, but here are some suggestions:

• make sure to generate request id on the server (if it is not already done by Runtime). One idea could be to simply reuse the activationID, that would make queries in splunk easier ? (if we are ok to expose it, otherwise a hash of it?)
• include the request id in every log instead of having a start and end log (as logs could come in mixed order for concurrent requests)

The request ID I think is already generated by I/O Runtime, we can add activation ID to this if required. It will have a comma separated x-request-id value something like nfjdjd34j454j4jd433, 433kdkdkddn5455wwdsfs. WDYT @moritzraho ?

Oh ok yes, just tested with an echo action and indeed there is a request id generated

so then ignore my first point, but second point is still valid

include the request id in every log instead of having a start and end log (as logs could come in mixed order for concurrent requests)