ace11, I don't know if you're the author of the video but in any case, thanks for posting a link to it.
It's indeed interesting.
We have been testing the sandbox with many malware samples and it usually did a good job at stopping them (i.e. isolated them). On the other hand, there are still weaker spots which we are, and will be, trying address to continuously (remember that the product has been released just yesterday).
But frankly, I don't believe the product will ever get to a state where it will be "perfect". I mean, I don't believe in silver bullets, and especially when it comes to security.
The sandbox is an extra layer of protection which makes great sense as it works differently than the other layers, and therefore has different strengths and weaknesses. It does a good job at stopping (or shall we say shielding) many samples that could be missed by the other layers. But no, it does not provide a 100% security (and whoever tells you their product provides a 100% security, don't believe him; all these systems work on a "best effort" basis, i.e. they're all penetrable).
Now, again, I'm not trying to discount the findings of the video, it's definitely worth looking into and that's exactly what we will do.
On a side note, don't you find it interesting how all those self-made tests on youtube etc. use Malwarebytes as a reference (e.g. to find out if the system got infected or not)? I mean, MBAM is certainly a nice product, but c'mon, we have literally hundreds of thousands of samples they don't detect.. which is not to say they're bad, just that it's somewhat strange to use their product as THE absolute reference in tests like this...