Several issues here: First, you need a non-greedy quantifier. And while we're at it, I'm sure you want something captured, so lets use the + quantifier instead of * (which permits nothing).

Next, you must also realize that your text has "newline" characters embedded within it. The . (dot) metacharacter excludes newlines by default, unless you use the /s modifier on your regexp. Here's a "repaired" version:

my $source = << 'END';
adf
<!-- InstanceBeginEditable name="guts" -->
this is the part we want
<!-- InstanceEndEditable -->
adf
<!-- InstanceBeginEditable name="crap" -->
adf
<!-- InstanceEndEditable -->
adf
END

# Pull out just what we need
my ($new_source) = $source =~ /<!-- InstanceBeginEditable name="guts" 
+-->(.?)<!-- InstanceEndEditable -->/s;
[download]

You're also not checking whether your regexp successfully matched or not, so when it silently fails, you simply get no text captured (in this example). You really ought to test whether a match took place or not.

One other word of warning: parsing HTML with a regular expression is fragile. You know, it's possible that a newline gets embedded in one of your HTML comments too, and that would break the regexp test. It's always advisable to use a HTML parser rather than rolling your own regexp approach.

my ($new_source) = $source =~ /<!-- InstanceBeginEditable name="guts" 
+-->(.*?)<!-- InstanceEndEditable -->/s;
[download]

my ($new_source) = $source =~ /<!-- InstanceBeginEditable name="guts" 
+-->(.*)<!-- InstanceEndEditable -->/;
[download]