A fluke.

You're using a key-value hash slice, which is used to obtain a list of key-value pairs from a hash.

$ perl -e'
   use v5.36;
   my %h = ( a => 1, b => 2, c => 3, d => 4 );
   say for %h{qw( a b )};
'
a
1
b
2
[download]

That's not possible in scalar context. In scalar context, it returns the value of the last key.

$ perl -e'
   use v5.36;
   my %h = ( a => 1, b => 2, c => 3, d => 4 );
   say scalar( %h{qw( a b )} );
'
2
[download]

In other words, it performs a simple hash lookup in scalar context. If you want to perform a hash lookup, it's much clearer to use a hash lookup instead of key-value hash slice.

$ perl -e'
   use v5.36;
   my %h = ( a => 1, b => 2, c => 3, d => 4 );
   say $h{ b };
'
2
[download]

In context, you used @{ %$hashref{ "To" } } aka @{ %{ $hashref }{ "To" } } aka @{ $hashref->%{ "To" } } aka $hashref->%{ "To" }->@* when it would have been clearer to use @{ $$hashref{ To } } aka @{ ${ $hashref }{ To } } aka @{ $hashref->{ To } } aka $hashref->{ To }->@*. There's already enough ways of doing this that it's easy to say it's wrong to add another level of complexity.

---

Update: s/hash slice/key-value hash slice/g, as per LanX's reply.