The third is close but it's even more likely that they just weren't prompted to survey at all rather than opting out. They're surveys, not automatic data harvesting, they don't represent things like Facebook's "total monthly active users". They're just a random sample of users, not a population count. Steam does sometimes report things like monthly active users in their annual reports, but haven't ever broken that down I don't think, but you can just infer it from the surveys.

I also have steam on multiple devices including a steam deck. On desktop I'm pretty much always logged in and I play games frequently, but most months I'm not selected for a survey. I use my steam deck less frequently and have maybe only gotten the survey prompt on it once or twice.

You REALLY don't need to care about querying the whole population for these kinds of statistics. Steam has a MAU of at least 132,000,000 users. To randomly sample from just 17,000 users (0.0012% of the MAU) would provide a 99% confidence the sample is within +-0.5% the actual value. Given that you've personally been selected multiple months, they are almost certainly well and above a sampling rate worth worrying about.

The survey does have other statistical noises/biases/errors though - but none nearly as large as this "gap". E.g. internet cafe reuse of the same machines by different accounts means if you're hoping for a "hardware popularity survey" instead of a "hardware users use popularity survey" then the number won't always make sense.