Follow-up to Dirty Contacts?
Yeah, it really does look like dirty contacts, probably on the 600.
I had some nice cloudy weather this morning, as opposed to that harsh sunshine yesterday, and I did some more tests on birds that actually show where the focal point is: ducks on water - the ripples often show the plane of critical focus or the width of the area between obviously soft focus zones, front and back. The latter is often easier to see and I look for the width of the sharpest zone to include the location of the single focus point that I almost always use. The bright cloudy weather allowed exposures at ISO 1600, f/11 and 1/500th second - pretty normal for my location over the greatest part of the year. Of course it gets a lot dimmer, but this is my default setting.
For analysis of focus accuracy and precision I used the lens wide-open, f/4 to f/8 depending on which teleconverter or none.
Working with a high-contrast test target in my back yard at a distance of about 40 feet (~12 m) yesterday I picked an MFA of -15. Today, working at varying distances up to about a couple of hundred feet (?m) I refined the setting to -18, but variation is so great that the exact number is meaningless. Also, note, that these subjects are 3-dimensional, not the flat plane target, so there's inherent variability of the spot the AF decides to select.
Sampling my saved image files (10-15% of the several hundred) showed considerable variation of the critical focus plane for easy subjects (ducks with good contrast and large features), just as I got yesterday in full sun. About 50% of the image files I consider as sharp as can be expected, the rest are just a bit off, but probably printable with some extra work. That is far better than the 1-10% I'd gotten previously, when the camera insisted on back-focussing, no matter what I did. And, again, closer subjects yeilded more images with better focus accuracy and wide-open aperture was useable, too.
In summary, what I think is going on is that there was a communication fault between the lens and the camera body which caused the AF to fail to recognize setting changes for micro-focus adjustments and that cleaning the contacts solved that problem.
However, it also looks like the 90d's AF system is only marginally accurate when used with my very long focal length lens and telextender combinations (1344 to 1920mm, effectively, and f/5.6 and f/8 wide open) , allowing the plane of critical focus to wander around the aimed point excessively, yeilding only about 50% critically sharp images under conditions that should have yeilded better than 90% critically sharp, as my other Canon EOS bodies have done. The 50% number is for good to excellent light and subject matter only, one would expect that the AF will perform much worse when even darker conditions or dark, low-contrast subjects get involved.
YMMV! Shorter forcal lengths are easier to stabilize and wider maximum apertures give the AF system more light to work with, so don't compare apples to oranges.
I am well aware that there a lots of other things that effect AF performance that are outside of it's ability to remedy, like atmospheric disturbance and the 3-dimensional shape of my subjects. These things have been accounted for in my analysis, as best I can.