Then there are the additional costs that are imposed. The threat model here is _very_ complex. Sprawling, even, with a lot of externalities.
You can mess up research results where that's at issue, you can accidentally encourage people to use these apps, etc.
You are unlikely to generate enough data for it to matter. Again: a single group by or filter term is not a huge burden.
So what can you do instead?
Time to get uncomfortable.
8/