When I was at Google, I worked in a tiny office, and I was friends with the recruiter who worked that office. A couple of times she asked me to vet the answers to the non-technical questions, because if the candidate went off-script in any way, she simply couldn't tell if it's a wrong answer or a more in-depth answer than the script anticipated. This wasn't part of my job, and if our office wasn't so tiny we wouldn't have had any contact, and she wouldn't have had any engineers to ask.
On the other hand, there’s some signal (sorry) on the nature of the work you’re doing in whether SIGTERM (15) is imprinted in your brain from near-daily appearance on your screen, or not.
I would say a vital command-line skill is having a good filter to ignore what's irrelevant.
Signals 9 and 15 and much more well known by their respective value than their symbolic name. They are everywhere in program output and log files and if those things are your daily work then you recognize this whether you want to or not.
It's like if you have touched computer code in any language then you probably know that 65 is A. Not because you have useless trivial memorized but just because it figures every time you look at string values for some reason. In one sense it is useless trivia, but on the other hand a programmer who never stumbled on ascii codes must have lived in an unusually insular bubble.
I have never written a program that outputs a signal number to stderr or a log file. And grepping the logs on a server with nearly 800 days of uptime, I can find only five references to signal 15, from months ago.
> It's like if you have touched computer code in any language then you probably know that 65 is A. [..] a programmer who never stumbled on ascii codes must have lived in an unusually insular bubble.
20 years of programming and doing things like making custom fonts. I couldn't remember the ascii code of 'A'.
A bit more relevant would be what's the difference between this two
So you're left with bad questions which have single answers, mostly involving rote memorization.
Additionally, it's common for interviewers to interview for multiple different groups (as an additional tech interviewer, a xfn interviewer, a Googleyness interviewer, etc), so exposure to job requirements can be reasonably broad at the interviewer level.
As I imagine many suspect, calibrating interviewers is hard, and always has been, for a company hiring roughly 10,000 people every year.