This will work fine if the input data can be completely relied on to be in exactly the form you expect.

Two things can take a chunk out of your posterior: (a) the data is variable (in any way) -- anything generated by humans is suspect (of course), anything generated by systems written by humans may not be perfect (let's face it); (b) you didn't fully understand the problem. More insidious yet, nobody may notice that the output of this program is incorrect until something REALLY BAD happens.

So, I'd do something like:

   if (/^LOGGING: \S+?-(\S+) \S+: (\S+) [^:]*: (\S+)$/) {
     my ($month, $what, $user) = ($1, $2, $3) ;
     if ($what eq 'STARTING') {
       $flushes{$month}{$user}++ ;
       $flushes{$month}{TOTAL}++ ;
     } ;
   }
   else {    # Any error reporting you like, eg:
     chomp ;
     print STDERR "Line $.: '$_' ??\n" ; 
   } ;
[download]

This is not a lot of extra code. Of course, you could do much more validation of the input and/or be less strict about some things (eg white-space and case)

Who knows, you might be relieved when you discover the file contains:

  ....
  LOGGING: 31-Mar-2008 23:59:10: ENDING ON A HIGH: countryjoe
  LOGGING:  1-Apr-2008 00:01:22: STARTING FLUSH: gotcha
  ....
[download]

before the auditors.