The fact that many of them are rotations/reflections of others also makes it much harder to recognise when text is in its correct orientation. Not really a problem if you're reading a book, but for things like signage and labeling I could see it being a source of ambiguity.

Looking at the glyphs really makes me appreciate the fact that traditional, "evolved" alphabets are naturally quite redundant --- it adds an important layer of error-resistance.

You think? I've spent a few days memorizing the script to the point where I can read roughly like a grade schooler (I don't see words yet, just letters). Slowly working through Dickens's "A Christmas Carol" [1] and I'm surprised. It's not a substantial problem.

The only place where I struggle with this is a few of the vowels, particularly around A sounds. That's because they're all short glyphs.

A much bigger obstacle is that I don't actually know how to spell most words phonetically. I've always had an excellent memory for characters, so I pretty much learned to read by memorizing whole words. So when I go to write a word like "today" and I struggle to choose between "𐑑𐑵𐑛𐑱" and something silly like "𐑑𐑵𐑛𐑧𐑘" (because there is supposed to be a Y there dangit!).

[1]: https://shavian.weebly.com/uploads/1/0/2/1/10212142/_a_chris...

I've been interested in making a custom script for a while but I always get stuck wondering if there is any research on most easily distinguishable shapes. Still not found any.

This is generally okay for natural language which already has a bunch of redundancy and doesn't need more. For non-natural language words, that's more important.