2400mm f/25 isn’t going to resolve significantly more than 1200mm f/12.6 by the time increased optical aberration and diffraction have kicked in, and will be horrible to handle. You would do just as well by buying Topaz Gigipixel and upresolving 2x.
Agree that 2400mm f/25 is not useful on a modern high-resolution body. You should consider that the 150-600mm zooms essentially have a 1.4x TC built-in to get from 400mm to 600mm, so just putting a TC on it is essentially stacking TCs (not exactly, but close enough to practical reality).
What I was getting at was that if you double up again the focal length of 150-600mm f/6.3 with a 2x TC on it already at 1200mm f/12.6, you increase the number of pixels by a factor of 4 but you hardly improve the resolution with a high resolution sensor and may even make it worse as TCs do degrade images to greater or lesser extents. So, you might just as well increase the number of pixels in post processing by using decent software like Topaz Gigapixel. I don't understand what you mean about ethical issues. Topaz Gigapixel doesn't do anything to alter the image artistically. All it does is to upresolve the image without making it softer. How is that unethical - it's no different from downresolving an image or sharpening it using software, and AI software does it better?
AI doesn't up-resolve the image--that's not possible. The original scene micro-detail is irretrievably lost when it is down-sampled through the Bayer matrix filter on the sensor. AI takes cues from images in a training set, and applies them to the image to simulate a higher resolution, creating false (but believable) detail. It effectively creates a photo-composite, only at the pixel level. So if you use it on a photo with a bird as the primary subject, it will use feather detail from birds in its training set to produce feather detail that "looks" good. If your bird photo is presented as abstract art, that's fine. But if you are presenting the photo as an accurate representation of that bird in the wild, it's not. If that's too abstract, consider whether it would be ethical to use AI to up-resolve a crime scene photo used in court.
What is your evidence that it is using training on a set of birds to create detail? Do you know what the neural networks are trained on? Are they just working from local general patterns? I can see in the patterns from the Topaz AI sharpening that they are doing a local geometric based sharpening and not reconstructing real feather detail.
From the Topaz web site: "Gigapixel AI’s neural networks are continuously trained on a wide variety of sample images to understand how photorealistic detail looks. After learning from millions of images, our AI image upscaler can now believably increase resolution on a wide variety of different subjects..." There probably are some feathers in those millions of images. Also note that they say it's "believably" increasing resolution. That's telling.
What AI is particularly good at is seeing patterns. If there are hidden patterns in interpolating pixels that it detects from say comparing low resolution images with high resolution then it can predict with some credence the best interpolation in various circumstances and not necessary using a database of say feathers. I think that Topaz doesn’t use accurate preset libraries of details as it makes a real mess of introducing artefacts when the settings are too high.
On my list of things to try is the RF 1.4x behind the EF 85/1.4L IS for a 120mm f/2. That's really about the only non TC-compatible EF lens I still have worth putting a TC behind. Double-stacking might be fun with the MP-E 65, I don't see a need for it with the TS-E 17 or 24.
As far as I'm concerned even the first TC is too much for MP-E ... with R7. With low res FF bodies 1.4x would be usable, but I seriously doubt 2x would help compared to 1.4x. The effective aperture is just too small especially as most MP-E:s tend too need some stopping down for max sharpness.
I would agree with that, though I haven't ever tested a 1.4x TC with the MP-E 65mm. The EF 2xIII TC adds some real resolution to images with the 1D X, but not with the smaller pixels of the EOS M. Nor do the smaller pixels of the EOS M add spatial resolution at 5x or 10x (but the smaller pixels do add spatial resolution up to 2x). That's all based on empirical testing at f/11, so wide open might be a little better but as you state the MP-E is not at its best wide open.
