That seems to be most studies these days. And 10 years ago it was mostly "we've recruited 200 students from [university performing the study]" and people complained about that too. And if you do a phone survey, then its "you're biasing toward the people who pick up calls from unknown numbers". There's always going to be some sort of sampling bias, and this study doesn't seem to be any sort of special outlier.
I don't agree, I think we'll get pretty good at this.
I agree there currently is pretty much always a sampling bias, though. Maybe the exception is when the study samples a particular group already, rather than the population at large.
I think in some fields this may be more obvious. If particles are too big for the LHC to find them, then we won't get meaningful data about those particles from those experiments. We can point out that we built a ridiculously, unbelievably huge collider and what do people expect? It's a valid point, but too bad for us.
(This definitely seems like a great idea and nothing can possibly go wrong)