Thank you for preparing this PR for issue #1753! Please can you add all additional test cases mentioned anywhere in the discussion for #3133 just so that we can easily be sure none of the problems with that approach are surfacing in your new PR? Much appreciated.

@gwhitney I had previously added the empty array and standard 1D cases, and I just added all the ones listed in this comment. I think that should cover it

Thanks! Yours was the first correct implementation for the 1d Kronecker product, and it showed how the 1D product could be factored out of the 2D product, which in turn made it clear how to support arbitrary dimension of the arguments (recursively). I've now added a commit that implements all that, and corrected the doc test for kron which was failing (you should do npm run build-and-test before issuing a PR, sorry if the contributing docs don't mention that?).

Except for needing a HISTORY line (reviewer's responsibility, not yours), this is good to go -- except unfortunately it is a breaking change (alters kron([1,2], [1,3]) from what was previously documented), so it needs to be scheduled for v15. I will try to update the v15 branch and merge it there; we shall see, I may need Jos's help.

OK, I had to rebase on v15 to get the HISTORY straight, and since I didn't know if you would want me to force-push to your fork, I created the replacement copy #3461 of this, which I am closing this in favor of. Thanks for the contribution!

Thanks @Delaney!