Arne Luenser

```cpp opentracing::Tracer::InitGlobal({}); ``` However: `Tracer::IsGlobalTracerRegistered()` will incorrectly return `true` after this. Should probably be fixed (and simplified).

Wouldn't returning a null pointer (default-constructed `shared_ptr`) have exactly the same semantics?

I tried out the branch mentioned above, and everything seems to work fine? Any reason why this could not be in 3.7.3? Maybe as an undocumented or beta feature?

This is implemented via token webhooks.

Indeed, the behavior is wrong for OAuth2 logouts with remember=false. This needs to be fixed in code; the only workaround right now is what @mig5 said [here](https://github.com/ory/hydra/issues/3132#issuecomment-1470872860).

This issue makes sense to me and feels like it should be fixed. Currently, the list of redacted headers is hardcoded [here](https://github.com/ory/x/blob/e1d7bd3cb462d841458541c7b2fecc070b373030/logrusx/helper.go#L62). We would need a mechanism and API in...

What's wrong with firing several requests instead? You can also do it in parallel if latency is the issue. I'm wary of adding APIs like this because they end up...

This is in fact a consequence of changes which landed in Hydra 2.2. Is this causing real world issues? The full flow still cannot be completed twice.

> Just to clarify, if both "password" and "generic" hook sections are present, then the generic section hooks still trigger for non-password flows, correct? That was the design goal, but...

@BrandonNoad I accidentally edited your comment instead of replying, sorry. Restored.