The first place I’d send you would be the module help, which has some descriptions of what each one means; a member of the team here also dug up this tutorial, which goes into a lot of detail on what each thing is and how it’s calculated with lots of example images.
All that being said, you’re absolutely right that unlike shape, a lot of the features like texture, granularity, etc are hard to put into “plain english” or even think about conceptually- it’s easy to come up with hypotheses for what might make a cell bigger, but harder to come up with ones for what might make a cell have a higher “Texture AngularSecondMoment”. FWIW, in our own lab we tend to not use those features individually that way but instead try to just incorporate them into an overall “fingerprint” or “signature” - we may not (yet!) understand what each feature individually contributes to the fingerprint in a given experiment, but it allows us to make really quantitative comparisons between treatments- ie the fingerprint of my mutant of interest is 20% similar to my negative control but is 90% similar to my positive control. This approach is called profiling; a labmate of mine wrote a great blog post on it a couple of weeks ago which links in turn to some review articles and a NatureMethods “best practices in profiling” that people in the lab here have contributed to. It’s worth checking out if you’re interested!