Another way, and configurable, but uses Perl 5.10 regex extensions:

c:\@Work\Perl\monks>perl -wMstrict -le
"use 5.010;
 ;;
 my $too_many_consec = qr{
   (?(?{ index('0123456789', $^N) < 0 && index('9876543210', $^N) < 0 
+}) (*FAIL))
   }xms;
 ;;
 my $min = 4;
 for my $s (qw(
   123 321  2123 2321  1232 3212 21232 23212
   1234 4321  21234 34321  12343 43212  212343 343212
   12345 54321  212345 454321  123454 543212  2123454 4543212
   )) {
   printf qq{'$s' -> };
   print $s =~ m{ (\d{$min,10}) $too_many_consec }xms
         ? qq{too many consec: '$1'}
         : 'consec free'
         ;
   }
"
'123' -> consec free
'321' -> consec free
'2123' -> consec free
'2321' -> consec free
'1232' -> consec free
'3212' -> consec free
'21232' -> consec free
'23212' -> consec free
'1234' -> too many consec: '1234'
'4321' -> too many consec: '4321'
'21234' -> too many consec: '1234'
'34321' -> too many consec: '4321'
'12343' -> too many consec: '1234'
'43212' -> too many consec: '4321'
'212343' -> too many consec: '1234'
'343212' -> too many consec: '4321'
'12345' -> too many consec: '12345'
'54321' -> too many consec: '54321'
'212345' -> too many consec: '12345'
'454321' -> too many consec: '54321'
'123454' -> too many consec: '12345'
'543212' -> too many consec: '54321'
'2123454' -> too many consec: '12345'
'4543212' -> too many consec: '54321'
[download]

Extended Patterns used are  (?(condition)yes-pattern|no-pattern) and  (*FAIL) (see Special Backtracking Control Verbs).

Updates:

1. Actually, the upper limit of 10 in
       m{ (\d{$min,10}) $too_many_consec }xms
   is not needed, and this works just as well written as
       m{ (\d{$min,}) $too_many_consec }xms
   instead. The unneeded upper limit might even be regarded as a potential future pitfall: what if you extend this approach to alphabetic consecutive runs and forget to change the limit?
2. Another afterthought: The regex object  $too_many_consec is misnamed. The "too many" function is fulfilled by the lower limit  $min of the counted quantifier; the regex following the capture group only tests for (and fails in the absence of) consecutivity in the capture, and so should be named something like  $consecutive instead.