I have no idea what the cost per unit translates to for eye-control. The research and development costs are presumably already embedded in the R3. So we may be talking a modest incremental cost to add the current system to the R5.
When I first got the R3 I was very excited to try the eye-control autofocus. After several months of trying to use it to shoot sports I shut it off and pretty much forgot about it. I believe there are two major problems. First, it doesn't react fast enough to follow action accurately. By the time it has found the subject the player has move on or passed the ball to someone else. Second, when I'm shooting sports my eye is not fixed on a specific spot, but is moving around scanning the scene, trying to follow the action and anticipate what is coming next. That rapid scanning causes the eye-control focus point to bounce around seemingly randomly and it is always a fraction of a second behind my eye.
I don't know if my experience is typical, but I have noticed a couple of things on internet videos (I don't pretend to have conducted a thorough disciplined survey). A lot of the videos demonstrating how well it works are shot under fairy controlled situations. For example, studio and portrait shooting or "action" shooting (usually done specifically for the demonstration and not under real world shooting conditions) where the photographer is only following a single person, such as an individual runner or rider. It seems to work well under those controlled situations, but honestly, why would anyone even need it under those conditions?
The other thing I've noticed is that when I've watched videos from bird and wildlife photographers who have been using the R3 for awhile, few if any mention the eye-control autofocus and don't generally discuss it in their "setup" discussions.
I'll readily admit, I'm just one person. But, if it was such a great feature, I would expect a lot more people would be singing it's praises and posting videos showing how well it works.
I think Canon will go one of two ways. They will either invest millions more to perfect eye-control autofocus or they will leave it as is and just include it as a feature on their R5, R1 and R3 lines with incremental tweaks every generation but not really investing the kind of resources that would be needed to make it a truly stellar feature. I'm guessing they'll go the second route because the investment needed to make it truly effective could be prohibitive.