I suggest you crunch your sample code down to a simple runnable test case that demonstrates the error. Something like the following (which does not reproduce the error):

use strict;
use warnings;

my $critical_failures = _analyze_test_logs ();
print STDERR "Count of critical failures: $critical_failures->{count}\
+n";
if($critical_failures->{count} != 0){ # line 183
  print "Condition true"
} else {
  print "Condition false"
}


sub _analyze_test_logs {
    my $critical_failures = 0;
    my %critical_failures;
    $critical_failures{count} = 0;
 
    $critical_failures{count} = $critical_failures;
    if($critical_failures{count} == 0){
      print STDERR "The count of critical failures is 0.\n";
    }
    
    return \%critical_failures;
}
[download]

Prints:

The count of critical failures is 0.
Condition false
Count of critical failures: 0
[download]

Update:

There are a few foibles in your code. $critical_failures->{count} != 0 is probably better written as $critical_failures->{'count'} != '0'. In particular the != is numerical but '0' is a string - how do you really want to compare it, as a number or as a string?

Using the same identifier for both a scalar and a hash in the same scope leads to confusion at best, and near impossible to track bugs at worst. I'd be inclined to change $critical_failures to $critical_failure_count.