I realise it's not an issue now, but it seems likely that calling keys in scalar context will reset the each iterator in (very small) constant time. Indeed a small test confirms this:

use strict;
use warnings;
use Benchmark qw(cmpthese);

my %bigHash   = map {$_ => undef} 1 .. 1e3;
my @resetEach = keys %bigHash;
my ($key)     = each %bigHash;
my $count     = keys %bigHash;
my ($key2)    = each %bigHash;

print "First key '$key', second '$key2'\n";

cmpthese (
    -1,
    {
        scalarKeys => sub {scalar keys %bigHash},
        listKeys   => sub {() = keys %bigHash},
    }
);
[download]

Prints:

First key '559', second '559'
                 Rate   listKeys scalarKeys
listKeys       5184/s         --      -100%
scalarKeys 15969899/s    307964%         --
[download]

My original assumption was that keys would have to be called in a list context if I was to extract anything useful from it, and that calling it in scalar context would still cause a full traversal in order to count the key/value pairs.

As usual, it turns out Perl is a bit more clever than expected, so using keys to reset and each to pick a pair may indeed be the Right Thingtm even for a huge hash. Maybe perldoc -f each should point this out more clearly?

-- Time flies when you don't know what you're doing