Any one help with a statistical question on the comparison between ONS infection survey rates (static) versus test results (growing) for Covid-19?

As ONS make clear, sampling frame mean that the infection survey is not accurate for small areas. But reversing that logic does that mean that the national ONS number will be less accurate in a situation where outbreaks are localised?

My intuition is that you need to have a large enough sample that you have a good probability of sampling in the localised clusters. And that means you probably need a much bigger sampling frame than usual to get a national number.

In spatial econometrics, the parallel would be the need to adjust significance levels a lot when your errors are spatially correlated.

A little rushed so perhaps this is silly or obvious but I couldn't find any discussion of it on a quick pass. Any pointers?