One of the things I have to keep reminding myself of is that the people claiming that non-response bias isn't that much of a problem anymore are the people whose careers depend on non-response bias not being seen as a problem.
"we'll weight the responses to control it" sounds okay on the surface but it really means you make a dozen giant assumptions about turnout models and they're even riskier because you have low n backing data. If you knew that you'd already know the results.
<2% response rates are insane. Useless.