1 - https://gridjumper.files.wordpress.com/2010/10/picture-5.png
I have a Vive dev kit and it is amazing, but I agree the amount space is going to be an issue. However, since Oculus and Sony aren't going to support the same freeform movement, there will be plenty of experiences that get ported to Vive that don't need a lot of space either.
I haven't found gorilla arm to be an issue; most interfaces seem to be going with laser-pointer style where you don't have to have your arms out. It's more like a gyromouse. I've had no issues with pretty long sessions inside Tilt brush, which is probably the most demanding in that regard.
Several new locomotion techniques are coming on line from all the experimenting that are solving the sea-sickness issue without having to dedicate an entire room to physically walking around. Here's one of the most promising ones: https://www.youtube.com/watch?v=lKnM5gC-XpY , seems to have pretty much fixed things for sensitive people.
The initial Oculus release with gamepad only and no motion controllers is going to be a bit of a let-down; you are right that just playing the same games seated with a gamepad or mouse and keyboard in VR probably isn't going to be enough for it to be a mass phenomenon.
Attach rates for hardware add-on to a video game console are historically horrible. They might be lucky to get 5%.
And for gaming, it's always about the killer app. Given that we're still in a learning curve for VR game development, it's not unreasonable to say there won't be one next year.
The main way I see it not happening is if it doesn't make it out in time for holidays with enough in stock.
Oculus has already shipped >200k dev kits. Assuming Oculus and Valve/HTC ship at least a million (>5X the number of enthusiast/dev kits), the 3 million total would have to mean Sony only gets a <5% attachment rate. Not likely.
Sony's headset will be competitive to with the Rift recommended PC specs (a GTX 970), even though it has way less power: the PSVR is primarily targeted at 60Hz reprojected to 120Hz, and it has a 1080p display that is RGB stripe instead of the Rift and Vive's pentile-esque 90Hz 1200x1080x2 displays. These two combine to mean PSVR will only have to push half as many pixels, for similar quality levels (PSVR actually has more subpixels than Rift and Vive; they throw away 33% of the subpixels you spent gobs of GPU power rendering). You aren't going to play games with fidelity of console TV stuff, but you aren't going to at the Oculus recommended spec either.
He doesn't explain why he thinks it will take until 2018 for Sony to get into the millions. We know the PSVR is design is finalized. As long as it is out significantly before Christmas 2016, they should get into the millions much earlier than 2018.