I still think kanji input is a bad idea. Showing the kanji after enter is pressed is fine, but kanji input via a typical Japanese IME actively sabotages the learning process.
For example there’s a grammar point for の下で “under”. If kanji input is accepted, we can type at least した, しも, or もと to write the correct kanji. If bunpro couldn’t verify that we correctly type もと, in my mind it would be a failure as a learning tool. It would reinforce the wrong answer instead of correcting us.
It happens at lower levels too, e.g. there’s an N5 grammar point for 好き. Because this is often read ずき I can type that too to get my phone to suggest 好き. But in the test questions it’s never read ずき and bunpro should not accept and reinforce it.
Accepting kanji is like accepting fuzzy answers or answers with typos. It’s like manually marking answers as good because in one’s mind the answer was “close enough”. It breaks the SRS.
I’ve previously also pointed out that kanji suggestions when using a Japanese input method can give away mistakes too without giving bunpro a chance to score accordingly; for example when we know the kanji but it’s not showing up in the suggestions for some reason, we can infer we made a mistake. IMHO this is more of a self-discipline thing but it’s something to be aware of.