To be fair, I know a fair number of people who are good at competitive programming but are absolutely awful at writing maintainable code.

Yeah, those competitions are not really representative of actual skill.

I was into competitive math as a teenager and was somewhat successful, but I actually kind of suck at math.

Similarly, I'm a professional developer but I'm really bad at competitive programming: what usually happens is that I know how to solve the problems but the time limit is too low (for me, at least).

I'd say success in competitions is a good indicator of dedication and perseverance, but not sufficient to spot someone who's good at the job.

Yeah, competitive programming forces you to use short variable names and write makeshift code which is fast enough to pass all the test cases ...

Because you can determine if they organize and test things in a repeatable and maintainable way or do they have trouble organizing structures and make questionable performance decisions. Are they clear on the hasA vs isA. Do they know what a mutex or static scope is? These are the things that will cause huge debugging nightmares. Syntax issues are no where close to as problematic so why use whiteboards vs an actual computer? In my experience of interviewing, questions about Security and Threading(performance / micro-opts) are good for separating the wheat from the chaff.

By the same way we have doctors do surgery in place, construction workers do a toy house, teachers give a class for free, cooks spend one day serving meals for free,...