I’m pretty sure that’s due to Unix’s early existence on PDP machines from DEC, which used octal instead of hex for encoding binary values. Why DEC did that I don’t know - I never got a chance to use DEC equipment back in the 70s.
Several PDPs used 18 bit word sizes (we standardized on 8 bit bytes and multiples for word sizes later on). 18 is not divisible by 4, but it is divisible by 3. An octal digit groups three bits together, just like a hexadecimal digit groups 4 bits together.
