Maybe I'm wrong, I'm a very causal user, but I think the only way to do it is to create a shape that covers the part you want to keep, and then do object -> set -> clip or something like that; it's not exceedingly awful, but it's not as smooth as clicking "crop" and dragging borders.
> Also, there are not 2 workflows, one for rectangles and one for arbitrary shapes, which is a good thing.
I think having a unified workflow is a good thing, and that that can exist along side important special cases.
A tool like inkscape gets a ton of "casual" once-in-a-while use and some amount of frequent power-user use. It's tricky to facilitate both kinds of use, but I think that should be a goal.
What you wrote is not ridiculous like the following statement, but consider "Yeah, you have to draw by using your text editor on an SVG file. There are not two workflows, one for drawings and one for text, which is a good thing."
You could "just" crop by subtracting shapes manually every time, too. Rectangle crops are extremely common so it makes sense to have a tool for that even if it just generates the underlying concept(as with shape-to-path).
If I were approaching the problem fresh, though, I would look into the presentation of the modal functions as the root of the issue. The source of a lot of complexity in image editing software is confusion over layers, masks, groups, and selections - they have some conceptual overlap and can be applied modally(e.g. shape as rendering element OR mask OR selection). If the UI buries the difference it just adds to the confusion.
