I can't say I agree with the overall approach of trippledubs here, but I think the use of prototypes therein can be said to have a constructive and well understood purpose.

As written,  collectLeaves() is recursive, and within the function the calls such as
    &collectLeaves( $_, $leaves ) for ( @{$ds} );
serve the purpose of suppressing prototype checking so the  $leaves reference can be passed in recursive calls.

Outside the function, prototyping allows calls like
    my @wanted = grep { $_ == 7 } collectLeaves $hashref1, collectLeaves $hashref2;
to be made, while
    ... collectLeaves($hashref1, $hashref2) ...
will fail to compile.

c:\@Work\Perl\monks>perl -wMstrict -le
"sub S ($) { return $_[0] + print 0+@{$_[0]}, ' elements in ', 0+$_[0]
+; }
 ;;
 print S [], S [42], 33;
"
0 elements in 7450732
1 elements in 30219348
74507333021934933
[download]

(But  S([],[42]) will not compile.)

Again, I don't say I necessarily agree with the approach, but I cannot agree that prototyping has been sprinkled thoughtlessly onto the code as a kind of magical correctness dust, as we have often seen. OTOH, I agree with you that the response of future maintainers to this code (or recursive functions in general) is a valid concern.