Quick ugly hack known to apply to https://github.com/jrockway/cperl-mode.git, but probably applies to whatever version of cperl-mode you have as well (I don't expect most people like messing with that chunk of voodoo code). Turns off parsing of multi-line command blocks. In your example, formats the first 5 lines in pod face, and the rest as code.

Update per Laurent_R's request: (not sure what level of explanation you were wanting, so forgive me if I go on too long) POD treats all text following an =WHATEVER as part of that "Command Paragraph", so

=head1 Heading Text

Blah Blah
[download]

=head1 Heading
       Text

Blah Blah
[download]

Both make an H1 containing "Heading Text" (it may or may not include a newline in the second case / I'm not sure and isn't important in this discussion). The Emacs syntax highlighter (cperl-mode) was highlighting LanX's subroutine as though it were an argument to the =cut. Apparently, perl itself knows that =cut doesn't take an argument and executes the line after it, thus executing the subroutine. The code below is a diff against the emacs syntax highlighter (cperl-mode.el) that comments out its parsing and highlighting of command arguments. To use it, you would need to download a copy of the cperl-mode source code, apply the diff using patch, then load the modified cperl-mode.el in your emacs configuration file. Before the patch, emacs would color =head1 red and Heading Text blue in the examples above (in my color scheme, probably different in yours). After the patch, the whole =head1 Heading Text is just red the "Heading Text" argument is not parsed, the whole line is just treated as POD. This makes the =cut span only a single line and emacs highlights the subroutine as perl code, as desired. The patch removes a feature of the cperl-mode parser so that LanX can get his work done, it isn't an improvement of cperl-mode that would be desirable generally.

diff --git a/cperl-mode.el b/cperl-mode.el
index 00e6c3b..d0b23ee 100644
--- a/cperl-mode.el
+++ b/cperl-mode.el
@@ -3873,38 +3873,38 @@ the sections using `cperl-pod-head-face', `cpe
+rl-pod-face',
                  (put-text-property b e 'in-pod t)
                  (put-text-property b e 'syntax-type 'in-pod)
                  (goto-char b)
-                 (while (re-search-forward "\n\n[ \t]" e t)
-                   ;; We start 'pod 1 char earlier to include the pre
+ceding line
-                   (beginning-of-line)
-                   (put-text-property (cperl-1- b) (point) 'syntax-ty
+pe 'pod)
-                   (cperl-put-do-not-fontify b (point) t)
-                   ;; mark the non-literal parts as PODs
-                   (if cperl-pod-here-fontify
-                       (cperl-postpone-fontification b (point) 'face 
+face t))
-                   (re-search-forward "\n\n[^ \t\f\n]" e 'toend)
-                   (beginning-of-line)
-                   (setq b (point)))
-                 (put-text-property (cperl-1- (point)) e 'syntax-type
+ 'pod)
-                 (cperl-put-do-not-fontify (point) e t)
-                 (if cperl-pod-here-fontify
-                     (progn
-                       ;; mark the non-literal parts as PODs
-                       (cperl-postpone-fontification (point) e 'face 
+face t)
-                       (goto-char bb)
-                       (if (looking-at
-                            "=[a-zA-Z0-9_]+\\>[ \t]*\\(\\(\n?[^\n]\\)
++\\)$")
-                           ;; mark the headers
-                           (cperl-postpone-fontification
-                            (match-beginning 1) (match-end 1)
-                            'face head-face))
-                       (while (re-search-forward
-                               ;; One paragraph
-                               "^\n=[a-zA-Z0-9_]+\\>[ \t]*\\(\\(\n?[^
+\n]\\)+\\)$"
-                               e 'toend)
-                         ;; mark the headers
-                         (cperl-postpone-fontification
-                          (match-beginning 1) (match-end 1)
-                          'face head-face))))
+;                (while (re-search-forward "\n\n[ \t]" e t)
+;                  ;; We start 'pod 1 char earlier to include the pre
+ceding line
+;                  (beginning-of-line)
+;                  (put-text-property (cperl-1- b) (point) 'syntax-ty
+pe 'pod)
+;                  (cperl-put-do-not-fontify b (point) t)
+;                  ;; mark the non-literal parts as PODs
+;                  (if cperl-pod-here-fontify
+;                      (cperl-postpone-fontification b (point) 'face 
+face t))
+;                  (re-search-forward "\n\n[^ \t\f\n]" e 'toend)
+;                  (beginning-of-line)
+;                  (setq b (point)))
+;                (put-text-property (cperl-1- (point)) e 'syntax-type
+ 'pod)
+;                (cperl-put-do-not-fontify (point) e t)
+;                (if cperl-pod-here-fontify
+;                    (progn
+;                      ;; mark the non-literal parts as PODs
+;                      (cperl-postpone-fontification (point) e 'face 
+face t)
+;                      (goto-char bb)
+;                      (if (looking-at
+;                           "=[a-zA-Z0-9_]+\\>[ \t]*\\(\\(\n?[^\n]\\)
++\\)$")
+;                          ;; mark the headers
+;                          (cperl-postpone-fontification
+;                           (match-beginning 1) (match-end 1)
+;                           'face head-face))
+;                      (while (re-search-forward
+;                              ;; One paragraph
+;                              "^\n=[a-zA-Z0-9_]+\\>[ \t]*\\(\\(\n?[^
+\n]\\)+\\)$"
+;                              e 'toend)
+;                        ;; mark the headers
+;                        (cperl-postpone-fontification
+;                         (match-beginning 1) (match-end 1)
+;                         'face head-face))))
                  (cperl-commentify bb e nil)
                  (goto-char e)
                  (or (eq e (point-max))
[download]