Most stack languages i've seen have the occasional bit of prefix hanging around things like definitions or flow control.
Yeah, I noticed that Fif uses prefix for function defs, but I figured that might be a pragmatic decision.
: Function (comment) function body ;
The colon, name and stack comment are all in prefix position. But in an interpreted environment you could maybe swap the semicolon to prefix and the name to postfix but you're still stuck with some prefix token to prevent execution.
Forth also has
Bool if code else code which is essentially prefix. But it boils down to the same reason, you need some prefix to delay execution anyway.
From a sort of conceptual level this is all fine anyway. A big part of stack languages is how they lead to a data flow sort of programming rather than some fundamentalist belief in postfix syntax. So when there is a language construct that doesn't benefit-from/contribute-to that data flow style people choose syntax they find comfortable or readable which is sometimes prefix or even infix.
For instance, in the repl functions, why is it "see <function name>" not "<function name> see"?
Most of the repls I've used use the same function invocation syntax for built in repl functions as for language ones, although obviously this doesn't have to be the case.