I bought the Samyang 24mm cheap second hand. (most of my gear is second hand). I did not like this lens, it had very low contrast. The 24mm Nikon is on another planet.
I tried shift lenses, back in my D300 days. the 28PC was wasted on the smaller format, as it became about 42mm eq, to long for architectural photography.
Sadly you need to use FF to get the full potential out of these lenses, for Architecture.
That can be subtle in a small displayed image, because of the circle of confusion, but a large print or projection will make the difference in resolution obvious when people walk up close to it.
We're talking about buildings here, mainly, with repeating patterns of bricks and windows that don't respond well to resampling issues.
I've never owned a tilt-shift either; I usually choose to zoom out wider than I would if I were just fitting the building in the frame with the lens pointed upward, to simulate a shift with a smaller sensor. That doesn't mean, however, that I don't recognize that a shift-able lens of just the right AOV would be the best possible option. Using all of your sensor and getting composition done in-camera is almost always the superior option, and having the sensor parallel to the building face means more consistent focus down that face, without having to rely on "DOF", which actually thins, visually, with high image magnification.
All lenses loose sharpness in the corners, my shifted 24PC is better than some ordinary lenses in the corners when shifted. My Laowa 15mm shift lens looks better at full shift, than my Z14-35 corrected in post. With all my shift lenses, even the ancient Nikon 28 and 35, you have to pixel peep to see the corner problems. The same is true of most of my ordinary lenses too.
The Canon 17TS is a rather extreme TS lens. I have read a few reviews of this lens and I chose the newer Laowa 15mm TS because it is optically sharper. It is worth noting that the Canon 17TS needs to be serviced every so often as the optics decanter with use, according to a couple of user reviews. This probably explains the F8 F16 comment.
But again it is not about sharpness, it is about composing the picture.
I've worked reasonably quickly with my shift lenses, and even with my D800 I have used my 24mm T/S Rokinon without a tripod. Tilt, OTOH, always takes time, and trial and error too, and is much more demanding of a tripod, for me anyway, YMMV.
One thing I've learned about perspective correction in post processing (and I sometimes have to do that with a PC lens too) is that both the vertical and horizontal axis need correction. What I mean is that when you correct the vertical axis to straighten the perspective the horizontal axis becomes too narrow and a perspective adjustment of the height is required to avoid making the scene/subject look unnatural. Thus, whenever I squeeze (I never stretch) the top or bottom of the perspective I also grab both bottom (or top) corners and shrink them so that the proportions remain close to how I saw them.
I do like to take my time when I'm photographing a scene (versus a subject), but when I'm bringing my camera along and I'm with other people, which often happens when I'm traveling, there is generally little time to spend. When I'm by myself, I will take my time setting up and even more time waiting for the light – perhaps that's the difference between travel and landscape/cityscape (i.e., "deliberate") photography, at least it is for me.
I have shot hand held with my D850 using the LCD screen and the in camera level. I found it a bit difficult as I am used to using a viewfinder to shoot.
I have worked out a fixed routine when using my Shift lenses. I first get the camera level and get the focus and composition approximately right. Then I get precise focus using a magnified view and then close down for the shot. This speeds things up. On a tripod I use the LCD with both my D850 and Z7. For tripod shots, I prefer the D850, because it has a dedicated button for bracketing, which makes HDR sets easy.
As for post, it depends on the software. With Capture 1, I made some shots where I shot the same view with the same lens, both corrected with shift, and with the camera tilted, and corrected in post. I then made some overlays. They pretty well matched. I made a Blog post with the results.. When I used diagonal shift, the results were very hard to duplicate with keystone correction in post.
I don't see any comparison there to other options; just the detail of particular items at different shifts on the same lens.
The kind of softness that you get at the outer parts of the image circle is certainly noticeable compared to the center, but I don't think the losses at the periphery can be as bad, as say, stretching the top of a building shot with a tilted camera 2x - 3x, especially when the camera has a low pixel count, and never even recorded all the detail at the top of a building that the lens projected.
The lens in the link was a 17mm, which is extreme for a shift lens. The more modern Nikon 24 as well as my older Nikon 28mm and 35mm PC lenses are quite good on the edges. Stretching an ordinary lens will exaggerate the softness by 2 to 3 times, as well as exaggerating and barrel distortion.
It can benefit from a shift lens inasmuch as you have control of the degree of the parallel lines converging/diverging.
As to cropping in camera, no, that's not what is happening. In fact, you are able to avoid cropping and get more pixels in the shot, so potentially more resolution rather than losing that to cropping.
I believe the "cropping in camera" explanation is one of those simplified phrases that explains how a shift lens works. My 24mm PC is really let's say a 20mm lens.
My sensor just uses a smaller "cropped" part of the image circle, but I can move the sensor around to get the image I desire. Just like when people use an ultra WA and crop out let's say the bottom foreground.
My Pentax K-1 allows shifting the sensor to mimic a shift lens. The lens must cover the shift. There is less shift than a shift lens. I need to use it more.
Thanks,
barondla