They changed it to multi-point.
Another WAG - they may have been hoping to leverage the built in image processing capabilities of the iMX6 SOC. It is a pretty capable device in that regard, and probably explains why they opted for that particular device family instead of much cheaper options. Since the heavy lifting for motion planning is done in the cloud, they could have went with a much less expensive processor.
Sadly, it seems that this major selling point, like many of the other lofty claims, has been walked back.
I’ve been holding out hope that the team over there is just busy with developing GFv2, and that is why there is little progress on the promised features (continuous autofocus, highly accurate visual alignment, auto-alignment for flipped over material, passthrough alignment, auto detection of phones/tablets, etc.).
The low hanging fruit is ignored, and only updates of very little utility seem to creep out. This would lead one to think they are sandbagging, and planning to include all these wonderful enhancements in a new product.
Then, they go and do something like the autofocus speed up. I don’t seem to recall any torch bearing mobs storming GF HQ because the autofocus wasn’t fast enough. So, why expend developer time on it? And why roll it out in such a haphazard way that it ends up taxing your already overworked support staff, and further damaging your shaky QA reputation?
To be a fly on the wall in that organization.
BTW - it looks like shortly after we chatted about their mixed Glassdoor reviews, there are two new reviews up there (this one and this one, both dated Aug 30) that look suspiciously like they were written by a marketing person.