I did some more testing to try to figure this out because your reported rare symptoms are weird.

Much to my surprise, the preprocess routine in File::Find does affect the subsequent directory traversal in File::Find.
From the documentation, I didn't think that it would do that, but it apparently does.

The nodirs() sub will leave the initial starting results directory in the results to be used by get_files();

There is no need to use File::Find here because you are only looking at files in the base directory and not navigating to subdirectories.
I would consider just using opendir (my $handle, "dirname") or die "$!" and use readdir(). I am unable to simulate a situation to reproduce your symptoms on my current computer. An unlocked data structure would explain this.

From your code, it appears that this is a multi-process application. Right now, I am thinking that perhaps there is some rare situation where $BASEDIR/$dir doesn't exist.

#!/usr/bin/perl
use strict;
use warnings;

use File::Find;

find( { wanted => \&get_files, preprocess => \&nodirs }, "C:/test" );

sub nodirs 
{ 
   grep !-d, @_;
}

sub get_files 
{
   #return unless ( -f $_ ); #testing without this statement
   # prints:
   # C:/test <- this is a directory nevertheless!
   # C:/test/SomeBogusFile.txt 
   
   return unless ( -f $_ ); #testing with this statement
   # prints:
   # C:/test/SomeBogusFile.txt 
   
   print "$File::Find::name \n";
}        

__END__
Test Directory structure
C:/test
   SomeBogusFile.txt
   /subdirtest
      docinbsubdir.txt
[download]