I have been reading a bit about the Orion spacecraft. I see they just did some sea recovery testing. It seems like in the post Space Shuttle era, NASA is going backwards. Is this a one off solution, or is there a solid rationale for moving away from controlled flight reentry?
Please realize that capsules like the Orion fly a controlled aerodynamic re-entry. They are not dumb balls thrown into the atmosphere. They have low Lift/Drag ratio compared to winged air/spacecraft, but they use their lift nonetheless.
Wings for a precision runway landing are all fine and dandy, but they are simply dead mass for missions beyond Low Earth Orbit (LEO).
The same can be said for the wheels - unless you want a re-usable spacecraft, wheels are a nuisance.
Missions to Mars, asteroids, or the Moon intend to bring back the crew and samples only. Unlike complex spacecraft, samples are insensitive to high g-loads/shocks at landing. This leaves you with the safety of the crew as the ultimate factor in mission design.
Those missions will be infrequent and not exactly time-sensitive as far as search and recovery go. Thus, you can allow an hour or two before the crew is recovered.
The final choice (water or land) depends on the country and the structural limits of the spacecraft. Say, Russia historically had almost no aircraft carriers, a sturdy Soyuz reentry capsule design, and a large territory without too many man-made landing hazards - rivers, railroads, transmission lines, mountains, marshes (okay it's technically Kazakhstan, but whatever). Not so for the United States. Anyway, making safe landing into water is easier at interplanetary re-entry speeds (terminal braking solid rockets/engines also take away precious mass budget). For instance, the Orion lost its capability to land on the ground after a mass reduction campaign scratched away its air bags.
in the post Space Shuttle era, NASA is going backwards
This is based on the assumptions that because the Space Shuttle looked cool, it was (a) an effective spacecraft, and (b) an effective method of reentry. Both assumptions are entirely incorrect.
As a route to orbit, the Space Shuttle was significantly less effective than anything else. It was less robust and more expensive per launch than the Soyuz. It was way more expensive per launch than Ariane. Its initial design concept of being cheaply reuseable was entirely incorrect - it was more expensive to refurbish a Shuttle than it would have been to throw it away and build from scratch, because designing for "refurbishing" made it way more expensive. As a spacecraft, it was a complete dead end. Its only achievement as a spacecraft was to somewhat resemble an aeroplane, a requirement set by the USAF with no technical basis.
Not only was it a dead end, but it also caused some steps back in spaceflight generally. "Investment" in the Shuttle directly prevented many unmanned missions being run. Its long development time and overruns were the direct cause of the loss of Skylab, resulting in the ISS having to be started entirely from scratch. Its lifting capacity to orbit was also no better than Ariane or Soyuz, so building the ISS required many separate launches with relatively small components, compared to the much more capable Saturn-V rocket which launched Skylab as a single payload. It prevented development of any manned missions beyond LEO. And of course we have two major disasters over its operating life, the last of which left the US and its allies with no manned route to orbit.
Then we come onto its reentry abilities. The most obvious problem was with its heat tiles, which caused the loss of Columbia. Replacement of heat tiles is why it was so expensive to reuse, of course. But also its basic shape with wings was a massive design flaw for reentry, because the structure of the wing root is inevitably a weak point. As a safe way through reentry, it really wasn't.
And finally of course, the last few thousand feet to landing. The Shuttle famously was hard to steer, had a glide slope significantly worse than a parachute, and a landing speed of 200-300mph. It was so bad at landing that runways had to be specially built to handle its awful performance, and crews nicknamed it "the Flying Brick". Landing it was extremely high risk, because as a glider you only got one shot at the landing, and its performance meant that highly skilled pilots were needed to land it - which after all was why the USAF had specified that it had to have wings, to ensure their continued involvement in the program.
Now compare to a capsule landing in the sea (US design) or a deserted plain somewhere (Russian design). The capsule can be designed to "shuttlecock" to the correct position for reentry. The capsule heatshield is a continuous smooth surface, so there are none of the weak points of a wing root. Parachutes give inherent stability with no pilot intervention required, and with multiple parachutes you have built-in redundancy. In essence, the capsule and parachute design just works.
So NASA got it right the first time, and then spent 50 years screwing it up in the pursuit of something which looked cool and didn't actually work well. After billions of wasted dollars and one unnecessary disaster, they finally returned to something which just works. I don't see how that's going backwards.
Disregarding other concerns such as the practicality of spaceplanes:
If you are designing from scratch, and you just need to get a crew and/or a few hundred kg of cargo into space and back to the ground, and you are not trying to push for hardcore re-usability with fast turnaround, a capsule is vastly easier to design and to make safe and durable and robust than a spaceplane.
Most capsule-type reentry vehicles are controllable to some degree.