They aren't pushing features like video to their consumer lineups very aggressively, because they don't want to cannibalize the sales of their cine lineup (where a $50k body is almost entry-level). At one point, a lot of the innovation was happening via the Magic Lantern project, which is an open-sourced firmware replacement emerging from the community, and it seemed that Canon was (understandably) not really interested in helping.
In that classic echo of Canon moving into the electronic/autofocus niche that Nikon left underserved, Sony was perfectly happy to move into the video niche that Canon left underserved. Sony's MILCs started as niche products, but they've gotten better and better, their consumer cameras are 99% as good as Canon/Nikon's best but a fraction of the price, and their flagship stuff is just off the charts.
About 5 years back, Sony added on-sensor phase-detect pixels, which have put them into contention with DSLR-class products, if not actually better. For stills, they perform virtually on par with in-prism phase-detect sensors. They are much better than contrast-detect AF. For video, they can actually autofocus while shooting, which a DSLR cannot do (if the mirror is up, no light can get to the sensors in the prism to focus) and DSLRs must fall back to contrast-detect, which is quite slow.
https://www.youtube.com/watch?v=up8K_xd_iwU&t=628
On top of that, it's quite a bit cheaper to ditch the mirror box and viewfinder (or go electronic), and it lets you build a much smaller camera. DSLRs are fundamentally limited by their register distance (sensor-to-mount distance), MILC has no mirror so the register distance can be much smaller.
Canon's MILC products are hot garbage, they tried a token product a couple years ago but they were timid to avoid cannibalizing their DSLR and video lineups. So it was a terrible product, combining the worst characteristics of every system on the market. Small sensor, terrible AF, late to market, tiny selection of lenses. I haven't been following too closely but maybe they will make a more serious attempt next time. For the time being though, they have effectively ceded a niche that is growing in importance.
https://www.youtube.com/watch?v=LO7rxitFLZg
To put it in a nutshell, MILCs are now "Good Enough", and Canon has missed the boat. SLRs have inertia but they won't forever, there are really only a handful of places where they have a legitimate advantage, and those are fading quickly. Meanwhile, they are more expensive than MILCs for an equivalent amount of capability.
As for sensors, Canon may have iterated faster earlier, but they haven't kept up lately. Sony's sensors are flat-out better: higher-resolution, better high-ISO performance, PDAF pixels, etc. Nikon's products are actually technically superior to Canon's at this point, because they're using Sony sensors and Canon insists on using their own in-house slop.
Sony, meanwhile, has slapped their sensor on a shake table and now has in-body IS/VR like Pentax (which admittedly is not as good as in-lens VR, but you can always disable in-body VR if your lens supports in-lens VR). And their high-ISO stuff is just godly, we're talking about shooting at ISO 102k or 204k and getting a pretty decent image/video out of it.
https://www.youtube.com/watch?v=CVLBHMbRMW4
It's that classic story of a company that becomes too entrenched in its existing revenue streams and can no longer innovate out of fear of cutting its own throat... and then another company goes ahead and does it anyway, and the slumbering giant has no response prepared. Just like when Nikon let Canon take a foothold.
(edit: looks like Canon released a new MILC a month ago that is at least a reasonable effort, although it has some quirks, really lacks any unique/compelling features, and is about 5 years late to the party)