CR is a good thing, in effect, it allows layered radiography, but no matter what the latest bells and whistles are, the death nail of RT will not be in capability, but in security and cost. Ask yourself what kind of chaos would be created if some terrorist tapped up a Co60 source under a table at your local coffee shop, or tapped one to a pile of HE.
Security of those sources has and will become paramount as there are people actually thinking along those lines based on data I've read to date, and the change of rules between then and now. (01 to present)
That security is going to come at ever increasing cost. Radiography will probably never go away entirely, but it is losing it's prominence with every day that goes by, and will to continue to do so for the forseeable future.
The RT guys can jump up and down and throw a fit all they wish, but at the end of the day, the associated cost of security, extra training, licensing etc, will make it cost prohibitive for many of it's applications in the long run. Witness the move towards AUT for pipelines and tanks. They were once nearly the sole realm of RT, that is no longer the case.
As to advances in UT, they are coming hand in hand with advances in computation. As an example; Those advances in computation that spawns the phased array units in use today.
The theory has been around as long as the wave mechanics has been around, but cost effective implementation was not up to speed.
Phased array utilized the principles of constructive and destructive interference,
http://www.kettering.edu/~drussell/Demos/superposition/superposition.htmlby pulsing (phasing) multiple beam components which are combined, steered or otherwise interacted with in such as manner as to produce meaningful results.
Add this with multiplexing of many elements, you get sectorial scans, and in many cases, equal or better imaging than radiography.
Once again, all of this is possible through advances in computational science. With the electronic data, comes the ability to record that data. Most PA units come with that ability these days. Hall effect/optical encoders, can attach to the inspection head to record x,y locational data, that is correlated to responses on the screen much like AUT. In fact, there are PAUT systems out there that have the best of all worlds.
EMAT, and others are also a viable option, as are AET options.
I believe you can see the commonality. That being computational speed. The faster that gets, the more it can do. The same cannot be said of radiography.
There is a group of people out there hell bent on making RT survive, but in my opinion, they are fighting a losing battle.