Actually this sounds like a form of exe watermarking. I'd wager they're doing something like re-arranging the locations of the functions in the binary for each individual Steam user, for example. That plus lots of other techniques would have the same effect as schemes like SecuROM i.e. discouraging casual copying.
You're probably relatively close to the truth here. I remember Age of Empires 2 randomized values during a multiplayer game in the page file that it kept to prevent online cheats, but the problem was that the "randomization" really wasn't very random, could be predicted, and if memory serves there wasn't any other cheat prevention mechanism, giving way to rampant online cheating.
The problem with such a watermarking is if they use any kind of "formula" for randomization, it'll be predictable, and thus can be broken. Then again, avoiding the online portion activation/checking in with servers nullifies the whole point, but at that point you can't patch, upgrade, or connect online.
If I had to guess, what you'll see is a combination of your watermarking idea, with impulse/steam's ability to download wherever you have a steam account. I expect they'll rely more heavily on online activation, delivering the exes online during activation, to help limit day one warez, with a checksum of the exe bound to your account. That way, they can track *who* leaked the copy of the game that eventually became warez.
This still won't stop the warez scene, but it might put a dampener on 0-day warez, which is where the serious fight seems to be focused on. Most rational devs have flat out stated that they know piracy will never go away, and focus on limiting it in the first 60-90 days, with a focus on 0 and 1-day releases.
Interestingly enough, if a dev environment was set up properly, *every* exe throughout development could be custom-watermarked and tracked during the dev cycle, and thus, over the long term, could track the individuals who work inside the industry who get hold of the game and leak it to the cracking groups for 0-day releases. I'm sure there's a relatively small number of them who are responsible.
I suppose even that could be circumvented though by modifying the exe extensively enough by the crackers. There might be enough forensic evidence though left to determine the source though, depending on how it's done.