I am not a Machine Vision expert, but my experience with Machine Vision cameras is that they provide uncompressed (RAW) data. The file sizes are HUGE and aren't realistic for a security application.
The lack of compression is what allows these cameras to run at such high frame rates even at multi-megapixel resolution.
So even if these cameras were able to send compressed data the frame rate would likely slow down.
Also, the transport method is important here. Many of the ultra high frame rate cameras use USB3 (like the PointGrey camera above from what I can see), which is not normal for surveillance infrastructures. Moving to GigE (Gigabit Ethernet) you may see lower performance on the frame rate side.
How many machine vision DVRs or encoders are you aware of?
How many HD-SDI DVR's are there?
HD-SDI outputs are used by many machine vision cameras, and they work with HD-SDI DVR's. Here's one that Gige, HD-SDI and ONVIF.
But even if there were none it doesn't affect my point. Matt is specifically referring to the infeasibility of using Machine Vision cameras with security applications in general, due to the HUGE size of the files.
My point is that HD-SDI security cameras stream uncompressed RAW files to HD-SDI security DVR's without issue. Therefore, just because a camera outputs huge, raw, uncompressed video does not make it unrealistic for security use.
Btw, here is a Lilin 120fps marketing video. Take a look:
To me, this is actually evidence of why 120fps is not needed. The 30fps video results, even with a money counter, appears to deliver enough frames to show the money movement clearly. The 120fps is obviously better, but it's not clear to me what practical difference that would make.