After some careful consideration, I went back into KiCad and replaced the VCNT2020 by a TLV4946-2K. This means that I will use a hall effect latch instead of the optical sensor I initially planned to use. I did not plan to use this sensor as I got tired after sifting through a few pages of hall effect latches on mouser only to find them relatively huge or slow.
The size is an issue as I want to keep a total thickness of about 8mm and I am already stuck with a 1mm wall supporting the battery. Of course, I could go a little thinner, but I would rather not. Thus, all through hole parts are instantly out of the question, while a lot of regular SMD parts are as well.
Slow sensors are also not an option since a sensor with 50Hz or 3000rpm means I would have no room to spare. Since I would really love everything to work even with compressed air, that's not really an option either.
These considerations initially lead me to select a very thin optical sensor and call it a day. However, there was on issue that kept me thinking: optical sensors draw quite a bit of current. That also means that my dream of using the index sensor to wake everything up would not have been possible.
So today, I went back and sifted through a few more pages of datasheets until I found a sensor that fit my bill. At up to 15kHz it's a lot better than the ones meant for detecting a closed display. I might even be able to use multiple magnets as described by [hyperneurone] in the comments.
I already updated the files on github and will now try to come up with a circuit that allows me to disable the LDO from software until the spinner is spun. That way I could get great standby times.
P.S.: I already ordered the parts that take longer because they are not available at Mouser. In this case, that's the two most important parts: ESP8285 and APA102-2020.