There are a lot of factors that impact on what is in focus: achieving a sharp image is a collaboration between the optical precision of the lens, the focusing system of the camera and its configuration, and the skill of the photographer.
As you admit, shooting through a window, which is a chunk of glass with no real quality control from an optical point of view (was it single pane or double glazed?) is not conducive to accurate results. The factory camera lens configuration is done under clinical conditions where temperature, humidity, air movement and light are all controlled: yours are not. Thus, in the real world all of these will change and they impact on the image your lens will capture. But unless your lens is grossly defective, by far the greatest variation in sharpness of image will be from how your focusing system is set up in-camera, and how you use it.
Simply relying on the focus you have adjusted is also not a constant - Bill (Waddizzle) has already made the suggestions for how to consider two settings for very close and very far. If this seems crude, there are reasons for this. Canon do not make a docking station for their lenses, because they make lenses only for Canon cameras, while 3rd party makers design for multiple brands, thus a Sigma lens will not have the facility to do in-camera micro-adjustments and uses the method you may have seen on You Tube, for example. Canon don't need such a device because they can provide adjustments in camera and they can be configured by camera firmware.
This is done in-camera for JPG files by in-camera correction software - you will find an item in the focusing menu set to have that turned on - it usually is. However, when you bring your RAW images into post-production software, that software itself will make adjustments to the focus and a set of other configurations according to algorithms provided by Canon, or created by Adobe, for example.
I was shooting a couple of days ago with that same lens and watched the images change as the lens corrections were applied to the RAW images in post. This is normal practice for modern lenses. However, you could compromise those corrections by trying to anticipate them - lens corrections are best made at the extremes of focus and then let the correction algorithms do their job. In all my four decades of photography, I have never felt it necessary to micro-adjust a lens - and my standards are high too as I used to sell my images.
Right now you are using the excellent EF 100-400L MkII lens, which is a relatively heavy, expensive unit because it relies predominantly on glass to get optical clarity. With the newer lenses across all camera makers, there is now a trend to create much cheaper, lighter lenses that rely less on purely optical perfection and now use computational photography algorithms developed by AI to perform dynamic mathematical image corrections - which can be altered or improved by firmware updates. This is particularly true for lenses at wide angles, where distortions and vignetting are the biggest issue, and that resulted in lenses traditionally were massive, heavy and cost a lot of money. One example is the excellent RF 14-35L f/4 lens. Without corrections it has massive distortions and vignetting, but with corrections on, it is tack sharp and square. Again, those corrections are turned on by default in camera and applied automatically in most PP software, and the results look excellent with them applied.
As I said, how you focus will have a huge impact. Default settings will tend to focus on the closest object within a specified focus area. The wider the focus area the more objects that encloses. With wildlife that is quite often not the subject's closest eye, which is the critical element. Thus, I have been shooting single-point centred Back Button Focus to get absolute precision of focus location. I focus on the eye, lock focus (which is always in servo mode), recompose and shoot. If the subject is moving, I hold the AF-ON button to track, if it is static I just tap it to lock it.
I used to have the 5DIV but have sold it, but in the last couple of days was shooting hand-held, in available (dim or contrasty) winter's light with the EF 100-400LII on the even more demanding 52MP 5DsR, and these are some of the results. The insect was shot through glass at the zoo, while the bird shots of the wood pigeon were in dense, dark bush. Image shot RAW, corrected using PS auto corrections, adjustments made for lighting levels, cropped and massively downsized to post here.
