I don't think a new iterator is the best generic approach.

Because only the loop's body knows if resetting the iterator is appropriate.

for the general case - iterating a hash and occasionally deleting elements in the body, this should always fit

while (my $key = each %hash)
{
  if ( condition($key) ) {
      delete @hash{ dependent_keys($key) };
      keys %hash;
  }
}
[download]

The Rule: reset the iterator if and only if other elements are deleted.

This will never end in an endless loop, because the hash will

• either be emptied ²
• or the last "re-iteration" is not interrupted.

• If the $key itself is deleted
  => this should be done outside the if-block to not reset °
• If the $key itself is NOT deleted
  => you might want to avoid processing it again by a checking a %seen hash.

my %seen;
while (my $key = each %hash)
{
  next if $seen{$key}++;
  if ( condition($key) ) {
      delete @hash{ dependent_keys($key) };
      keys %hash;
  }
}
[download]

°) not sure how costly resetting is in the end.

²) provided it's a finite structure which isn't refilled simultanously