Events + traces = observability.
They are not necessary, and often not even /usable/, for observability.
Whip that shit into shape with some structure and ordering already. Again, orthogonal to observability unless you turn them into events and gather lots more context.
Literally nobody.
We have metrics and logs for legacy reasons. Solid, real, true reasons -- but all of them legacy.
So you need to be able to slice and dice by user requests, with as much context and descriptive detail as possible. (Traces are just derived from ordering fields in your events.)
To say nothing of all the ways they fuck you by preaggregating on aggregates at write time.
NOT THIS TIME