I figured that if Brad Abrams, Kryzysztof Cwalina, and Joe Duffy all agreed that you call the async callback before you signal the wait handle, then they must be right. Sure, blocking on the wait handle in the callback results in a deadlock, but perhaps (I reasoned) that is not a supported use case.

Joe refers to a customer who emailed Brad and Kryzsztof about this. That’s me. Then I backed off, figuring that Joe’s original post shed some light on the issue. I’m not usually so quick to defer in this way, but at least I deferred to some really bright guys.

Note: I turned off comments because this post is a major recipient of comment spam.