Okay, deep breath. First of all, kudos to Lou for breaking this story. Go read what he has to say, and then come back for my thoughts. Good? Good. And now, for my take: sh*t! I wasn't able to run Lou's test on the pre-production Cosmos that I skied last winter because I'd already returned them. But my own, brand-new pair arrived today, and I can confirm I'm seeing exactly the same behavior Lou describes.
In fact, I wasn't even able to get my Cosmos to release from my Dynafit bindings. The force required to pop the boot free seemed to be getting dangerously close to damaging my bindings, so I stopped trying. Needless to say, compared to the far easier and smoother release of my Maestrale boots, this is an alarming discovery.
As Lou has noted, the lack of a brand-agnostic standard for tech-binding systems opens the door to considerable variations whenever manufacturers choose to eschew Dynafit-licensing. That seems to be exactly what has happened with Garmont's 2013 Cosmos. Even a cursory comparison to the real thing (ala my Scarpa Maestrales) reveals that the Cosmos' tech inserts are shaped quite differently.
Unlike 'official' Dynafit tech inserts, the Cosmos' are deeper and less tapered, with a sharper, steeper, angular edge. These topographic differences inarguably alter the retention and release characteristics of my Dynafit TLT vertical's toe unit, and least in comparison to my Scarpa Maestra boots. What is not clear is how relevant this difference is to real-world performance and safety. What is clear, however, is that they are different.
Garmont has just sent Lou an official response, in which Garmont's president asserts that tech system binding release values depend primarily on boot-heel construction and shell material, not the toe insert, and that in any case, during product development, Garmont extensively tested and was satisfied with the retention and release performance of the Cosmos' "Garmont Tech" inserts.
Lou has posted a short response to Garmont's statement, which you should read.
Unfortunately, I am not a Dynafit engineer, so I am not in an ideal position to make definitive comments about this issue, but I will note that Garmont's statement seems to categorically reject any possibility that something might have gone wrong between the development process, when their tests were conducted, and manufacturing of the current run. That strikes me as premature.
We should also be honest in admitting that tech-system bindings do not offer the same release and retention safety that a well-adjusted Alpine binding offers. I have long argued that skiers who take a lot of falls should probably not be riding on Dynafit or other 'tech' bindings. I have also argued that skiers should probably not ride Dynafit or other tech bindings extensively in-bounds—certainly not as their primary in-bounds rig.
That said, when I do ride a Dynafit or tech rig, I have the expectation that the binding and boot are working as Dynafit intended. It is difficult to believe that the very visible geometric differences in Garmont's tech inserts have no adverse effect on tech system binding performance. It is difficult to believe that the significant difference in force required to release the toe of the binding and the lack of re-centering behavior, as seen in Lou's test, has no effect on safety.
Nonetheless, it is possible everything Lou observed is actually irrelevant to real-world release and retention performance. That doesn't seem likely to me, but I frankly don't know enough about the workings of a tech binding to say for sure one way or the other. In any case, Garmont has made its position clear: the inserts are working as intended.
Unless Dynafit or Scott USA decide to weigh in, that leaves the final word on this subject up to you.
Note: my original review of the Cosmos is here.