Heath Stewart

Changing title based on discussion. Recording for separate frameworks is one way to do that, but there's some pretty significant downsides to that. We should consider all possibilities. One other...

True, but from what I've been told, much of that time spent doing management plane testing can't really be helped. FWIW, tests themselves can be parallelized if possible. NUnit defaults...

Maybe not thread safe on recording but I'd argue we shouldn't be recording in parallel. Should be thread safe to play back, though. I do it all the time and...

That said, this wasn't even about parallelism. Floating point de/serialization differs across .NET Framework and .NET nee, "Core". Recording separate files was one way, but I probably shouldn't have made...

Separate recordings per TFM is something I've been wanting to drive after some discussions late last year with @JoshLove-msft et. al. - basically, we record on all supported/desired TFMs and...

Well, CIs may actually take a little longer. Depends how much they are parallelized. But this would definitely and significantly reduce live testing times, which also reduces costs. Just want...

Do you have a sample of what that would look like?

Thanks, though, wouldn't it be `consumes`? It's part of the request header telling the server how much extra OData "stuff" to send.

@rloutlaw could you look into the "This Product" link at the bottom. It's 404'ing.

As for the change request, we could add a `DeleteIfExists` like we've done in other libraries including Storage. Newer libraries [shouldn't throw on delete by default](https://azure.github.io/azure-sdk/dotnet_introduction.html#:~:text=Deletes%20one%20or%20more%20resources%2C%20or%20no%2Dop%20if%20the%20resources%20do%20not%20exist.), but the pattern wasn't...