A study on mask effectiveness in schools wouldn't be one vs one, which I agree would be pretty poor science. With the proper design though, this is perfectly feasible and in fact is done commonly in other contexts. There is an established history of cluster randomised trials in education, where a subset of schools are assigned one intervention and another cluster gets another intervention, and you can then compare outcomes. In this context, I guess you'd choose a few local authorities and then assign some schools within each to each group (masks or not, or maybe some other arms too like voluntary masks), making sure you had a spread of schools to cover confounders like deprivation, and then see who fares best overall. (I don't design such trials personally, but I know they are done!)
The barriers to this are political rather than scientific, from unions and the FM not wanting to lose face or appear inconsistent, but that's not a good reason not to do them. And I do also believe that lab data is no substitute for real world effectiveness data - it's all very well saying they would work if only kids used them properly (for instance, I don't know if this is true), but if in reality kids don't use them properly then the intervention won't work.