Not a member there, but I'll comment on the technical aspects. While this is feasible in theory, there are too many variables that make it prone to false positives and effectively turn any collected "evidence" into circumstantial at best.
The seeder could be unconnectable...
or have the IP(s) used for the "random data block request" in its blocklist...
or not have good network connectivity...
or not have any free upload slots...
or have reached its maximum connections limit...
or be forcing protocol encryption in both directions...
or have randomized/changed its listening port...
plus other factors I'm likely forgetting right now.
Some may not remember or even know this, but IPT claimed to be using a similar method ("we don't care about speeds, we check how you respond to a BitTorrent protocol connection") to detect cheating as early as 2009, which was later found to be a complete fabrication. And besides... if you were tracker staff, would you announce your new anti-cheating measures and then proceed to explain in detail how they're going to work? Security through obscurity may not be a great policy, but it's a pretty effective one in a torrent world where 1. there is no presumption of innocence (if we think you cheated = you cheated and bye bye), 2. the value of what you're trying to secure is fairly low (in an era of fast connections and gigabit seedboxes, cheaters couldn't damage a tracker even if they tried), 3. the cost of unsuccessfully bypassing the security is high (losing your account and having to find another invite).
As for your last question, when RatioMaster Plus receives an incoming connection, it answers with a choke message (which basically means "I can't talk right now, try again later", like the no free upload slots situation mentioned above) in the same manner the emulated client would. Prior to this clever approach, all ratio tools including mRatio merely accepted the connection and sent nothing.