In may solution it is enough to query the keys of both hashes without ever creating the merged hash itself.

Sorry, but that is rubbish. This code:

my $joined_keys = 

[ keys( %{ { map { $_=>undef }  keys( %{$hash1} ) , keys( %{$hash2} ) 
+} } ) ];
# This.....^..........................................................
+..^ *is a hash* built from 
#...............................^^^^^^^^^^^^^^^^^...^^^^^^^^^^^^^^
#...the combined (potentially huge) lists of the keys from both hashes
+ PLUS
#.....................^^^^^...undef as the value for each key in the c
+ombined list...
[download]

And is functionally and memory consumption-wise, nearly identical to this code:

sub mergedHashKeys{ 
[ keys %{ {                           %{ $_[0] },         %{ $_[1] }  
+ } }  ] 
};;
[download]

And, your code creates *two (potentially) huge lists*; one going into the map and one coming out that is twice as long; from which the combined hash is constructed.

Any memory savings will be minimal.