The fixed IR sensor approach does have some advantages - it works in zero light level and you can see through the floor - but I doubt that they are worth the disadvantages of lower resolution and latency issues. At that point, you can go to a simpler (possibly uncooled) IR sensor for warning and tracking.
We're still very early in the design life of this technology. Latency seems to have already been resolved (at least enough to satisfy the testers) and I would think will only improve with time, assuming all else remains the same. But nothing stays the same, so as the resolution improves, the required bandwidth goes up with it, so latency will again become an issue. It will be interesting to see where this goes over time.