I don't think I've ever had to learn something very complex under the same short time constraints and high do-or-die pressure as an interview. Tiny things (think: new-to-me unixy commands), yeah, but nothing very complex. With as many unknowns as that introduces, should it ever come up in a real job situation with a very short time window available, my expert (I'm quite serious using that term here) advice would be that we should count on having it take quite a while, and immediately begin whatever mitigation is required to allow for that, rather than waiting to begin planning for and mitigating the harm that'll do and simply gambling that this complex thing won't surprise us and end up taking longer than we'd hoped (likelihood that it will: extremely high).
I end up doing a lot of technology selection and so I have to have these skills and the ability to ask hard questions.
I don't for a second believe this should be the skillset of every member of a team. A team where everyone is adding new tools all the time is chaos. Learning new tools isn't really a goal. Getting better at your job is the goal, and learning new tools is either a means to that end or just moving the goalposts over and over so nobody knows what's going on.
Yeah, sure, I do evaluative work just fine, but I don't think "build [non-trivial thing] with [fairly complex software or tools you've never used] while we watch you, you have 50 minutes" resembles that very much. The closest I've seen is situations along the lines of "meeting's in 50 minutes, could you find out about [thing similar enough to other things you've done or worked with that you are qualified to evaluate it on short notice]? I just learned we'll be talking about it in there and your take on [a couple specific aspects of it] would be useful." Not "you've never used this, I want a demo of [more than hello-world] in 50 minutes, or it's your ass", ever.
