I'm actually quite impressed by this choice. It's the one that gives the most flexibility in product dispersal. Also, I would predict that an .exe exporter will be written, just as soon as all the basics are down.
Flexibility? Somewhere in the future, maybe. Currently, HTML5 means you have to confine to casual games, because there is no hardware acceleration AND you can't tell or force, which browser the gamer will use. So, to let them be able to play within Firefox you have to be very careful, how many objects you use, etc. Comparing 2000+ sprites with C1 and 120 in C2 I feel this being a step back and not forward. If there would be an exe exporter WITH hardware acceleration (OpenGL, DirectX) and we had the choice, I would hail C2. But, together with Ashley saying, that there won't be a Scirra .exe exporter, I'm not very happy.
Hopefully this will encourage third party ports to new platforms. This could include an official Windows desktop EXE runtime in future, a third party Android native exporter, or whatever anyone thinks up.
Let's face it, HTML5 might be the architecture of first choice in 5 years - but I need something I can work with now...