Arpad Borsos

> I am currently initializing sentry (`sentry::init(..)`) _after_ I start the async runtime (tokio). Yes, you should switch that around. Sentry automatically inherits a per-thread Hub from the Hub initialized...

Ohhhhh, I think I completely misunderstood this from the very beginning 🙈 I was assuming that you wanted to propagate the trace headers to outgoing http requests you are doing...

> Is it more commonplace for the frontend client to start a Sentry transaction and generate an ID for a particular user session, forward that ID value in the API...

TBH, I wouldn’t put too much trust in them. In particular, I’m not even sure whether we are even testing the http interface at all, or just relying on a...

Thanks for the feedback and the idea. Indeed this problem comes up quite often, and I believe one reason is that `sentry` is split up into multiple crates, which in...

I think the problem here is that you are using `exec`. Per documentation, this function never returns, and the docs also state that this means no destructors (guards) are run:...

Finishing the span (and the root transaction) would submit those to the transport. Then you would have to flush the transport. I don’t know enough about your usecase to fully...

This should have been implemented a long time ago in https://github.com/getsentry/sentry-rust/pull/596. It might be possible that the transaction is not using the correct `Hub/Scope`, or the problem might be elsewhere.

This might actually be the case that we fundamentally do not support .NET minidumps.

The "observable" side-effect would be related to the timing of filesystem access. Already we advertise that users should flush any attachments in the on-crash hook, because thats when attachments are...