You're mistaking "representative" with "inclusive".

The survey can be a representative sample, without including your ̈́"circle of JS hackers". That's how surveys work.

Is there any reason to suspect Flow users are less likely to do the survey than, say, TypeScript users?

I can speculate here a little bit. Majority of survey users has less than 10 years of experience. That makes them susceptible of being Visual Studio Code users. VSC - Microsoft product. Typescript is Microsoft product as well.

I am curious (as a VSCode user) what makes you conclude that junior devs are likely to use VSCode?

It is really good tool and free one. Better tools are either commercial or requires a lot of learning.