Most proposals for 9 bit bytes weren't for adopting 8 bits of data in a byte, they were to have 8 bits for data and 1 bit for something else, typically either error detection or differentiating between control/data. Very few folks argued for 9 bit bytes in the sense of having 9 bits of data per byte.

9 bit bytes never made significant headway because a 12.5% overhead cost for any of these alternatives is pretty wild. But there are folks and were folks then who thought it was worth debating and there certainly are advantages to it, especially if you look at use beyond memory storage. (i.e. closer to "Harvard" architecture separation between data / code and security implications around strict separation of control / data in applications like networking.)

It's worth noting that SECDED ECC memory adds about a 20% overhead, though it can correct single bit flips whereas 9-bit bytes with a parity bit can only detect (but not correct) bit flips which makes it useful in theory but not very useful in practice.