As to being surprised at finding teams still that don’t use it, I’m no longer surprised at pretty much anything.
I’ve seen version control implemented in such ways that the teams would be better off without it, I’ve seen teams with no effective version control get by by using impromptu schemes like backing up the sourcetree in timestamped directories at regular intervals.
I’ve seen teams fail that had comprehensive version control policies because they were poorly designed policies that made the version control repository effectively useless (if there’s no way to find out which version of a file is in which release, having version control provides a false sense of security).

A lot of this seems like it might depend on the system being developed (how many images? how often do they change?). But it does seem like that’s generally a good idea.

I can understand your concern about maintaining versions of a web site, including the images on it. I have struggled with it, also. However, I have come to the conclusion that storing images in version control ends up being messy.

It is similar to the issue I have with external dependencies — having them in version control results in multiple copies of the same image or similarly named images all over the place. Then it is a manual process to view/compare images and see what images are where.

If a team/group can all agree on how to name images and can keep them straight, it may not be an issue. You will still have copies of images throughout your version control system, but disk space is cheap, so that’s not a big deal.

On the other hand, if a team/group can not agree on naming images (at the very least), I would propose another mechanism for storing images. Why not put them in their own repository that is meant for images? There are some good open source image repositories that allow you to tag and compare images appropriately. Also, if you use Adobe CS, you have access to Adobe Version Cue, which allows you to store, track and compare versions of images.

Then, your build script would reference the appropriate image from its repository location…

These are just my thoughts. I know there are other ways to accomplish this, but I hope you see what I mean.

Another really interesting point you brought up is not storing anything you don’t control the source of. I like this a lot. One point I might disagree with you on, though, is on images. Maybe I misunderstand the context you are talking about, but take for example a web application where you have a bunch of images. Say you don’t control the actual design of those images, but every once in a while, a designer or a business person says “let’s change the graphic on page X.” I prefer that changed image to be in version control with the change in source that points to the new image.