A bottleneck is certainly one issue. I saw the tests about applying a protocol filter, but they are flawed for a few reasons.
One thing is that a filter can't check inside achives with passwords, so there's an instant filter cercumvention method right there.
Or even using an encrypted protocol would do the job.
And unfortunately both of these things (and more) are methods that those sick people both know of and use, I'm quite sure :(