Most of this is implementation details that you don't need to worry about (and aren't guaranteed). The exception that comes to mind off the top of my head is that lexicals are indeed created for the first time before their containing block is entered, and you can set them via a BEGIN, INIT, or CHECK block:

{
   my $q;
   BEGIN {
      # runs at compile time when the outer block hasn't  been entered
      $q = 42;
   }
   print $q; # prints 42
}
[download]

However, changing it like so:

for (0..1) {
   my $q;
   BEGIN {
      # runs at compile time when the outer block hasn't  been entered
      $q = 42;
   }
   print $q || "undef"; # prints 42undef
}
[download]

shows that it does get restored to undef at the end of the block. Getting back to the original question in this thread, modifying it:

our $q = 54; # never actually used
for (0..1) {
   my $q if 0;
   BEGIN {
      # runs at compile time when the outer block hasn't  been entered
      $q = 42;
   }
   print $q || "undef"; # prints 4242
}
[download]

shows that keeping my from executing at runtime suppresses this clearing at the end of the block.