Let's assume I have a video of 10 minutes length. And I want to reduce its size by having some parts in high bitrate, while other less important parts can be low bitrate. For this I'd cut the original into segments. Leaving the important ones as they are, while transcoding the unimportant parts to a lower quality.

Now, is there a video container format out there which is able to "glue" these different parts together into one file? With good support in players so trusty mplayer will play the video as if it's an ordinary file? And in a standardized/annotated/undo-able way so a simple cli command would untangle the segments into separate parts/files again?

I was under the impression Matroska would be able to do this, via chapters, where each of my segments would be one chapter. (I've read it somewhere where merging videos without transcoding was dicussed). But when I read the actual mkv docs, it seems mkv chapters are more like a textual index pointing into another (video) stream of an .mkv file - and not some low level kind of data block index which would help me with my idea. So mkv looks like a dead end.

Simple editing, or merging (which is possible losslessly, with some remuxing) wouldn't offer me having different qualities in different sections of the video - right?

What I'm looking for is region-of-interest adapted quality, something image formats like JPEG2000 offer for spatial areas, only here applied to video over time.

Any suggestions or ideas?

The accepted answer suggests using the VBR capabilities of a format like mp4. While this works, please leave an answer if you know of a different way of achieving the desired result.