"Does the accuracy of Nikon's Z-system focus depend upon the amount of ambient light?"
It's taken me a while to confirm, but that's not really the case. The actual thing that determines how well the focus system is working tends to be the exposure under the focus cursor being used (including the auto subject detect ones).
There's a reason why Nikon prioritizes the matrix meter pattern to the subject detection (on the Z9 you can turn this partially off with Custom Setting #B4).
What's the difference between ambient light and exposure? Well, you can have low light present in a scene, but a correct exposure in the camera for that light. However, the exposure at the focus position is the one that will determine how well the camera focuses. If the subject under the focus cursor is well exposed, the focus system works fine. If your subject is backlit and underexposed—common in birds against sky situations—focus is going to be more problematic.
The Z9 is really the camera that started to confirm many of my suspicions about how Nikon is handling focus: the data is coming from the viewfinder stream, not the imaging stream. If the viewfinder is showing the subject as dark, focus suffers. If the viewfinder is showing the subject as bright, focus works. Even in low light. To some degree, the matrix metering works against focus on the Nikon Z bodies: while the overall matrixing generally produces a decent result, dark subjects that are backlit don't. And they won't focus easily.
Put another way: any underexposure of the subject that you want focus to be attained on will hurt the focus system's performance. That may provoke hunting, missed focus, or even inability to attempt focus in extreme situations. The Z system cameras actually focus quite well in low light, but only if the subject being focused is well exposed. On the Z9 Custom Setting #B4 allows the matrix meter to prioritize human faces in the exposure, but it doesn't do so for animals! Some of the contradictory statements you've seen about how the Nikon focusing performs has to do with inconsistencies in subject exposure, not focus system accuracy.
Some of you may remember that one way to deal with focus in low light was to either switch to a higher ISO or to move to a manual exposure that was intentionally set too high. In particular, the latter will help focus in low light, but then you have the problem that your exposure for the photo might be too hot. What we need is a button customization, Increase Exposure Temporarily, which we could provoke to get focus. Coupled with Back Button Focus technique, this would get you to focus quickly and reliably, but you could then let go of the button(s) for the photo to be exposed properly. I suppose you could do something similar with the second control ring set to Exposure Compensation, but because Nikon didn't put click stops on the second control rings, moving back to zero isn't exactly easy to manage.
"Should I use High efficiency raw on the Z9?"
My short answer is currently: no. The tendency of everyone is that they want to use High efficiency* on the Z9 because it makes the buffer less limiting. Of course, with the right card, I don't find the buffer limiting with Lossless compressed.
The problem with the High Efficiency settings is simple: Nikon hasn't done anything—other than putting in basic support in NX Studio—to make them useful to photographers. Meanwhile, IntoPix, whose intellectual property underlies the High efficiency options, isn't doing anything to give converter makers useful information, at least not that I can see. So we're currently in a situation where I'd say we don't have optimal raw conversion capabilities for High efficiency. Many converters don't support it at all.
Further, I've noticed two things with some early use of High efficiency: (1) I sometimes see slight color shifts, particularly at higher ISO values; and (2) noise, particularly in the shadows, isn't handled elegantly by the few converters that do support it.
Years ago there was a fairly large and organized group of users asking for "open raw" from camera makers. That dissipated when Adobe came out with DNG. However DNG wasn't embraced by Canikony, so we're back in the "we've got this great new raw format which we won't tell you anything about" situation. Simply put, the Canon, Nikon, and Sony "free" raw converters suck and are poorly maintained. I can crash NX Studio in a matter of seconds, and don't get me started on batch processing. The primary third-party converters we use to get the best images out of our cameras have to reverse engineer far too much with raw files these days, and don't always get that quite right. The lack of cooperation from the camera makers actually holds back what is possible in terms of image quality from their cameras. Go figure.
For the time being I'm sticking with Lossless compressed, because at least with that I can get to the most optimal data set and apply my own corrections using my preferred converters and other software tools. If you need more buffer space, set the camera to 12 or 15 fps instead of 20. My ProGrade COBALT card keeps up with those frame rates just fine.
And you videographers? You have the same problem, only in a slightly different form. Nikon's free "LUT" for N-Log isn't particularly accurate or well maintained, either. The Nikon/Blackmagic Design relationship isn't much better than the Nikon/Adobe relationship, so while DaVinci can handle the new Z9 8.3K 12-bit raw files, it's mostly incumbent upon you to figure out how to grade them well.
