We moved away from wrapping them recently for performance reasons.

How much of a performance hit were they causing?

and how many errors are they getting?

You got to add more context here. An entire blog post would be worthy. We do some very performant code at high scale and error wrapping has never put pressure on our systems. Highly interested in what you saw.

what programs are you writing where error wrapping represents a performance cost that's worth avoiding???

It seems like a waste of time to me. Wrapping errors adds context. But you can usually get enough context from stack traces.

But you don’t have stacktraces.