Add some documentation about missing Honeycomb spans

[dmvianna]

This PR adds a little Gotchas section to Core.Telemetry.Honeycomb to alert users that if they see a (missing) span in the Honeycomb UI, it may be just that that span hasn't finished executing yet.

[istathar]

I'm not entirely sure I agree with the sentiment

it is best to call beginTrace inside a function that will iterate continuously

but only because I think the best practices for doing larger processes (especially batch or asynchronous ones) are still emerging.

I might word this as "best to have a trace for each architecturally discrete service" but even that is only on the proviso that there is something non-deterministic between services that isn't worth measuring (or rather, isn't part of the behaviour of that [set of serivce[s]; you don't want to penalize the metrics for a service if a job waited for an hour in a queue).

Anyway, I'm not sure what's best to say here, other than "it depends" and what you've written is a great start. Merging.

[dmvianna]

About

it is best to call beginTrace inside a function that will iterate continuously

I left the root span below the iteration in a long lived process overnight. I just killed it in the morning. Now I have a single trace and no good way of finding which span gave me the error I'm interested in, and which I can clearly see in the graphs.

In short, in this case, putting the root high in the hierarchy was a bad idea.