There appears to be 3 types from my experience (now 6) and others' posted here and elsewhere:
1. Flat failure to get a lock (usually experienced with a small subject in midground)
2. Soft focus (most airtime has been about the R IV and 200-600)
3. The no-focus focus - the EVF or the EXIF shows a focus point when the actuality is fuzzy
Added in an edit:
4. 7R III fails completely to adjust AF to a sharp change in distance
5. A9 II doesn't acquire focus in AF Wide except in the centre square
6. Back focussing in the 400/2.8 and TCs (Update: this has been found to be a lens fault in three cases.)
The questions are how does Sony do hybrid AF and what's not working in these cases? I can only offer some ideas about the second question.
Sony said re the R II that AF started with phase detect sensing (PD) and completed the lock with contrast detect sensing. That makes sense (!). PD gets you into the ballpark with speed but less accuracy and CD is just the reverse (it's software analysis of contrast between neighbouring pixels and works by trial and error (except in Panasonics)).
Good light and contrast help both modes to work well and especially CDAF. If you have less than these CD takes longer and maybe simply runs out of time. The common zooms at full reach limit the light available quite a lot with apertures of f6.5 or f5.6. Add a 1.4x TC to that and there's still less light for AF to work with.
The bodies come with a settings scale of AF to Release priority and the default is 'balanced'. AF priority = don't shoot without sharp focus. Release priority = I want the shot regardless of sharpness.
When I select AF priority on the R III and A9 the fps rate with many BIF bursts drops noticeably. Down to 3 fps in a test I've just done on a setting of 10 fps on the A9. But not when there's a clear subject against a plain background.
This suggests that 'balanced' accepts soft shots. A hi res sensor like the R III and R IV is going to show that up more readily than a 24 MP sensor on the other Alphas.
It's also of course going to better resolve subject and camera movement, lens defects and heat haze.
Sony's PD sensors are line-type only which is less accurate than cross-type. That means their contribution to AF accuracy is lower and leaves more work to CDAF, which means more time is needed. And maybe the near-enough focus standard set by 'balanced' will let through soft images with low light/contrast subjects and/or slow focussing lenses and/or high burst speeds.
So the implication for sharpness is to manage these three dimensions yourself. Eg. slow the burst rate when the light isn't good (or go to AF priority). I can happily share with you that you don't need to spend squillions to get around the problem with a 400/2.8 prime because that's the lens I've been using (mostly with a TC).
Weakness 1 I've posted about elsewhere and I can say I still don't know why it happens. You can have Small Spot AF overlapping a static bird and the AF just reaches to what's behind it. Sometimes EXIF shows a focus point and sometimes there's none. I think it's an algorithm issue and it may be a result of CDAF kicking in too soon - CDAF tends to preference the more distant lock possibility (Panasonic's last f/w update for the CDAF-only G9 included a settings option to preference near or far).
Weakness 3 a number of us have seen occasionally in the EVF - get a green box but the subject it's on is fuzzy. In my AF/Release priority testing I found it happened twice with 3 shots each time having a fuzzy subject under content the camera recorded in EXIF as being in focus.
What I was doing was slowly panning across the scene, starting with mountains, crossing the intervening palm tree and continuing with the mountains. There was 3 fuzzy shots before focus changed to the palm.
The following screen dump is of one of these shot's focus point extracted by A7info.exe from EXIF.
The purple boxes are the PDAF points that were used by the body and the white circle is the focus point decided by the camera.
I've just read your other thread on talkemount.com and I have a suggestion about why your camera fails to acquire focus although with a caveat that I didn't see you say whether you were using AF-S or AF-C. The following applies to when the camera is using pdaf which I would expect to happen under most circumstances using AF-C. However, the same cause will also impact cdaf.
When a camera is using pdaf it's comparing two images from leftward and rightward looking pixels within the af point. The images are shifted with respect to each other for features which are out-of-focus. The camera can tell which feature in the scene under the af point is nearest and how far to adjust the focus to bring that feature into focus by the offset of that feature between the two images. However, this only works where the feature has sufficient detail (contrast) to enable the camera to match it in the leftwards looking and rightwards looking images. If the feature is so far out-of-focus that it appears completely blurred, the camera won't be able to "see" it in the af images. If another feature under the AF point does have lots of contrast, even if it is further away, the camera will focus on it. If there are no features under the af point that have sufficient contrast to focus on, the camera will rack the focus until there is. So, in short, with mirrorless cameras, pdaf can fail to acquire the desired subject (usually the nearest) when it is initially completely out-of-focus. There are two ways that you ways that you can try to get around this problem:
Pre-focus to approximately the correct distance. Of course, this in not always possible.
Zoom out to acquire focus and then zoom back in when you have it. Using a shorter focus length reduces the aperture diameter at a fixed f number and so increases the depth of field so more features are less blurred. Try that on your green pole example.
You may well ask why this appears to be less of a problem with dlsrs? As far as I understand it, this is due to the way pdaf works in dlsrs. Within the dedicated pdaf sensor of a dlsr, they use a mask to sample images from left and right sides of the image cone for an af point. They also have optics within the pdaf sensor that project those images at an aperture value of f/20-f/30 and as a consequence, far less of these images is completely blurred, so the camera can select the closest feature and focus on it more easily.
I'm no expert on dslrs but I read a long and interesting thread once about a guy who dismantled a pdaf sensor, took images of what the sensor "saw" with its line sensors and deduced how the camera used that information to calculate correct focus.
I had similar experiences to you when I first got my A9II and 200-600 and I am now constantly pre-focusing onto the mid-distance where I expect my subject to appear. It's not foolproof, and I still miss shots when I'm surprised by a sudden situation outside my expected range, but it helps quite a lot.
Just to be clear we're talking about complete focus failure (Weakness 1).
I found that AF-C appeared to work a little better than AF-S. Reading David Busch's guide to the R III, he says AF-C operates predictively. It may be that lacking a quick lock it does as you say and settles for the distant lock. I don't think it's switched to CDAF at this point since that's obvious by its slowness and hunting.
There are a few generalisations as to the subject matter I can make. They're tentative ones but we can test your hypothesis against them.
It may be more likely with a monochrome subject.
Sometimes I think it's less likely if AF is set to tracking.
It is less likely if the subject moves a bit.
It may be more likely if there isn't a good contrasty edge in the focus area but good light and a contrasty edge are not a guarantee of success.
It almost always happens with a kind of fence post we have here called a star post which is brown to black steel about 40mm wide say 5 to 10 m away (in my area a bird is often to be found perching on the wire near such a post but I can't use the post as a focus springboard).
So I think your hypothesis is plausible.
I have a couple of buts though ...
Sometimes the Small Spot is entirely on the bird and there's no more distant point to focus on.
Some of the completely failed focus shots' EXIF recorded no AF point but what's in the distance looks sharp or sharpish. So there may be an infinity default (when in doubt, rack it out).
It may be worth trying the middle focus range limiter and see what that does (on the lenses that have it). As a bird shooter swapping to that when presented with a subject will lose me shots, and I can't accurately estimate 5m or 7m or whatever distance away, but it may add to insight. (As an aside I'd like to see that setting and OSS mode software selectable and so assignable to a custom button.)
Some folk suggest manual focus but I don't find it easy with a 4.3 kg rig. It's slower than successful AF and loses shots too.
Re getting a kind of prefocus using AF, that's what I've been doing - finding something nearby as a springboard.
-----------
I'm going to add a Weakness 4 which may be related to 3.
Start on a midground subject in H or H+ burst mode and pan across to a more distant one or background; say from 5m to 20+m. The reverse works too.
The AF fails to adjust to the change in distance while recording shots but it does while doing a dry run.
I tested this on 3 different scenes.
The A9 under these circumstances appears to recover focus in a few shots (at 10 fps and AF priority).
I regarded this as a fault and Sony took it back and sent it for testing.
It was sent back with the assessment that they couldn't reproduce the fault. What the lab had done was change the AF/release priority to AF so the camera didn't record shots when the focus failed. No out-of-focus shots means no fault, right?
In practice this weakness isn't a big deal; my kind of bird shooting doesn't encompass these circumstances often. But it's part of a picture of AF performance.
I do not have the GM 400mm f/2.8 but I have not experienced any issues described here with my A9 nor A9 II using the GM 600mm f/4 lens. There are situations when I set the AF area to Zone-AF and the camera focuses on the closest contrasty spot within the zone rectangle. In that case, I switch to a Selectable AF with assists and so far so good. On a different occasion I tried to focus on some fox kits using the Center-AF but there were tall grasses nearby and the camera did focus on the tall grasses. I noticed it right away and I changed the AF area to Small Selectable and the AF of the camera was on target.
For BIF, I mostly use the Zone-AF and as long as the AF setting is not at the minimum focusing distance or at infinity while the actual target is 40-50 feet away, the camera grabs the focus and tracks it really well, too. I do prefocus prior to the actual action to a target that I anticipate the target would be, just an approximation. Most of the time the target doesn't change much anyway. In essence it is basically what @somersettr described above.
Not much of help here and that the problem you experienced is not real but what I described above has been my experience.
The only issues I really have are RIV/200-600 back focusing (which is essentially gone now with new FW and/or using OSS mode3) and failure to "see" a grossly OOF target closer to me than the previous plane of focus.
I believe that in AF-C there is no CDAF step. At least that was my understanding with the current cameras.
Oh and that Animal Eye-AF only works for mammals and rarely works for birds...but then Sony never said it would...just teased us with that Snowy Owl in the launch press event....but I guess to be fair it actually does work on owls and that is about it....
Focus for me with A9, RIII and RIV have all been very good and very consistent. Has it been perfect all the time? No! I have not had any flat failures like mentioned above.
