If one were to look at sales numbers, profitability and general trends in technology, a much better case could be made that full frame is the format that is heading toward oblivion.
Trend: Other Technologies – Software development is not standing still. Improvements in noise reduction, sharpness and resizing programming will further narrow the visible differences between APS-C and full frame.
I am not saying that full frame is dead or dying. But the truth is, if one wants to pursue this pointless argument, you need to look at all the trends, not just look at what camera manufacturers are doing as they try to capture a share of a relatively stagnant market.
I'd like to point out that the counter-argument to this particular argument is that it depends on Full Frame technology remaining stagnant. I believe it's a common argumentative mistake to assume everything else will remain the same.
Yes, cellphone and small sensor technology has improved vastly over the past few years, and it even rivals or supercedes P&S from 5-6 years ago, but how does it compare to a modern P&S - it is still inferior in most aspects.
Then there's the question of scalability, if a cellphone sensor is "good enough" can it be scaled to a larger format, and for most electronics the answer is yes.
So that brings us back to the question of APS-C and FF becoming obsolete, they won't become obsolete as long as the technology in them is updated.
Here's another example with something we're all using these days, processors. The current fabrication of processors has resulted in more and more transistors being packed into a small space - in fact the overall size of the processor is tiny in comparison to the 8086 of three decades ago. Now the shrinking of the processor is the fact that we can continue to make transistors smaller and smaller, once that limit has been reached we will be forced to make them bigger and bigger, physically; though technically speaking, if one were to only measure the size of a processor by transistor count it has gotten a bajillion times bigger. So because the physical limit, in the case of processors, is the size of transistors, we can make a better unit in the same size - but we've chosen to make them physically smaller because we have made them more efficient space wise because we can do more with less, this applies both to each actual transistor as well as size.
Now the problem with optical sensors is that by definition they're limited by size. It's not that we can't continue adding megapixels to cameras, the problem is with the optics themselves. The more megapixels we add, to a given size, the less forgiving the sensor will be to imperfections. Yes, we can fix the problems through software manipulation but that doesn't change the fact that you can do the same fixing on a larger unit.
Perhaps a more comparable analogy is car engine. These days, days we can coax power comparable to a V8 from yesteryear out of a V6. However a modern large V8, with the same technology as those turbo'd 6's, has more power.
The question really comes down to, will there be a demand for large FF sensors and if it's profitable or not for the manufacturer to produce them. Manufacturer have many ways to approach how to scale the technology:
1. They can physically just scale the sensor to a larger size, maintaining the same megapixel count - meaning that given the same technology, you'll have more sensitive/accurate pixels
2. They can scale the sensor to a larger size as a ratio - meaning you have a much larger FOV (crop factor) while maintaining an overall "same quality" as the smaller sensor
- Some manufacturers have done something similar to this through sensor stitching
As it currently stands, most manufacturers have taken some kind of middle ground.TL;DR Version
The advantages of a larger sensor will always be maintained as long as you aren't comparing different generations - and that's one of the biggest people make when they claim obsolescence. m43 is a great example of this, it has come a long way to challenge or even beat (depending how you look at it) the tried and tested Canon 18mp, but that sensor is 5 years old; that m43 scaled up to APS-C will beat the m43.
edit: I would be screaming "TAKE MY MONEY" if they fit a FF sensor into the SL1, which is my current dSLR. Looking at the mount, I see no reason why they wouldn't be able to: from a technological standpoint.