tracing across go routines

[gregwebs]

When errors are handled across go routines, can we ensure that gets traced?

In this case there should be multiple stack traces generated. Currently the code has some optimization to avoid generating multiple traces.

Generally code using errors is going to look for just the first stack trace it can find and then use just that- in addition to generating a trace per go routine we need at least an official way to coalesce the stack traces together.

[gregwebs]

Perhaps it is possible to leverage errtrace? https://github.com/bracesdev/errtrace
The problem with exclusively using errtrace is that you have to wrap errors everywhere to build your stack up.
With the existing approach of gathering the call stack, if that is done just once you get most of the information that you need even if there is no wrapping anywhere else.