POE FollowTail sometimes works sometimes doesn't

Hi,

I am facing a bizarre behaviour from Wheel::FollowTail.
Basically, sometimes it detects the log reset, sometimes it doesn't.
Sometimes it gets the record, sometimes it doesn't. From my extensive
tests, it seems to work only in 10% of the cases.
The file watched is overwritten by an application every 10-15 minutes,
contains only 3 lines. Here is an example of its contents:

   namp@tewap4 ~/code> cat /namp_log/log/TRLOGStatus.cfg
   #
   #Sun Sep 14 18:03:05 IRDT 2008
   TRFILE_INDEX_NUMBER=157

The file is on a normal filesystem:

   /namp_log on /dev/md/dsk/d50
read/write/setuid/devices/intr/largefiles/logging/xattr/noatime/onerror=panic/dev=1540032
on Tue Sep  9 00:58:53 2008

Any idea what could cause this?
O/S is Solaris, Perl 5.10 built from source using gcc, latest version of POE.
I'm about to revert to a more 'tradition' approach of watching the
file (e.g. checking its timestamp every minute) but I wouldn't mind
figuring out why FollowTail, which has worked perfectly fine on other
projects, behaves this way on this file.

Here is the simple code I am using.

sub InitWapLog
# Define POE records
{
        $poe_records{_start} = \&begin_watchers;
        $poe_records{index_record} = \&index_got_record;
        $poe_records{index_log_reset} = \&index_log_reset;
        $poe_records{log_error} = \&generic_log_error;
        &Log("Init complete");
}

sub begin_watchers
# Start POE sessions
{
        my $heap = $_[HEAP];
        my $logid;
        my $log_watcher = POE::Wheel::FollowTail->new
        (
                Filename => $LogStatus,
                InputEvent => "index_record",
                ResetEvent => "index_log_reset",
                ErrorEvent => "log_error",
        );
        $heap->{services}->{ $log_watcher->ID } = "index";
        $heap->{watchers}->{ $log_watcher->ID } = $log_watcher;
        &Log("Started watch on file $LogStatus");
}

sub index_got_record
# Handle index file reset
{
        my ($record, $wheel_id, $heap) = @_[ARG0, ARG1, HEAP];
        my $service = $heap->{services}->{$wheel_id};
        &Log("index: read $record");
}

sub index_log_reset
# An index log reset means the current log has been closed
{
        my ($heap, $wheel_id) = @_[HEAP, ARG0];
        my $service = $heap->{services}->{$wheel_id};
        &Log("$service log reset");
}

sub generic_log_error
# Handles log errors
{
        my ($heap, $operation, $errno, $error_string, $wheel_id) = @_[
HEAP, ARG0, ARG1, ARG2, ARG3 ];
        my $service = $heap->{services}->{$wheel_id};
        &Log("$service log $operation error $errno: $error_string\n");
        &Log("Shutting down $service log watcher.\n");
        delete $heap->{services}->{$wheel_id};
        delete $heap->{watchers}->{$wheel_id};
}

# MAIN

my $LogStatus = "/namp_log/log/TRLOGStatus.cfg";
&Log("Starting WAP log watch");
&InitWapLog;
POE::Session->create(inline_states => {%poe_records});
$poe_kernel->run();
&Log("All watchers died, exiting");
exit;