PHP - Server Segfault's When Trying To Pdo::rollback() ...

Hi folks, 

Running, 
PHP 7.1.4 
Lighttpd/1.4.32
on a small NAS, which is kinda old.... but all I've got as a server ;)

I'll try my best to explain my problem... hopefully someone can shed some light on this...

Essentially, I've written a "database wrapper" to handle any DB requests (selects, inserts, etc). It's using PDO...

At some point in my dbwrapper, I create a wrapperStatement object, passing along by reference, the PDO connection. After creating this wrapperStatement object, I call a function within it that'll do multiple inserts....

On my NAS, this multiInserts() function seems to be cause a 500 http server error. When I tried running the same script, directly on the server's command line with php-cgi, I get a Segmentation Fault. 

HOWEVER, I've installed PHP and Lighttpd on my working PC, and the code in multiInserts() just works !
 

class dbwrapper {
  ...
  $con = &$this->getConnection();
  $dbwrapperStmnt = new wrapperStatement($con)
  $dbwrapperStmnt->multiInserts();
  ...
}

class wrapperStatement {

  private $_conn;

  public function __construct(&$conn){
    $this->_conn = &$conn;
  }

  public function multiInserts(){

    // init data
    $data = array (
      array ( "jdoe1", "jdoe1@email.ca", "abc1"),
      array ( "jdoe2", "jdoe2@email.ca", "abc2"),
      array ( "jdoe3", "jdoe3@email.ca", "abc3"),
      array ( "jdoe4", "jdoe4@email.ca", "abc4"),
      array ( "jdoe5", "jdoe5@email.ca", "abc5"),
    );

    // init variables
    $username = null;
    $email = null;
    $passwd = null;

    $sql = "INSERT INTO users (USERNAME, EMAIL, PASSWD) VALUES (:username, :email, :passwd)";

    try
    {
      $con = new PDO("mysql:dbname=somedb;host=somehost;port=3306", "user", "passwd");
      $con->setAttribute(PDO::ATTR_ERRMODE, PDO::ERRMODE_EXCEPTION);
      
      //$con = $this->_conn;

      $stmt = $con->prepare($sql);
      $stmt->bindParam(':username', $username, PDO::PARAM_STR);
      $stmt->bindParam(':email', $email, PDO::PARAM_STR);
      $stmt->bindParam(':passwd', $passwd, PDO::PARAM_STR);
    }
    catch ( PDOException $ex )
    {
      die($ex->getMessage());
    }

    $errorList = array ();
    $totalRows = count($data);
    for ( $row = 0; $row < $totalRows; $row ++ )
    {
      $username = $data[$row][0];
      $email = $data[$row][1];
      $passwd = $data[$row][2];

      try
      {
        $con->beginTransaction();
        $stmt->execute();
        $con->commit();
      }
      catch ( PDOException $ex )
      {
        // HERE!!!
        // IF USING $this->_conn AS CONNECTION, SEGFAULT IS PRODUCED ON ROLLBACK!
        if ( $con->inTransaction() ) $con->rollBack();

        // just push the error to an array of failed transactions
        array_push($errorList, array ( $row, $ex->errorInfo[1], $ex->getMessage() ));
      }
    }

    if ( $errorList )
    {
      // print the errors
      foreach ( $errorList as $error )
      {
        echo "Row #:" . $error[0] . "<br>";
        echo "Code :" . $error[1] . "<br>";
        echo "Mesg :" . $error[2] . "<br><br>";
      }
    }
  }
}

Now, I've pin-pointed the exact line causing the segfault, and it's when I try a rollback() if I got errors (i.e. duplicate username, etc). 

What I've also found, if running on my NAS, is that if instead of using $this->_conn as my connection object (which I really should!), I create a NEW PDO object and use that one instead, it works, no segfaults!! 

At first, I thought it might have something to do with passing my $con has reference to my wrapperStatement constructor (perhaps I had done it wrong?), so I checked with var_dump($con)...

Before creating my wrapperStatement constructor, inside my dbwrapper... 
var_dump($con) = object(PDO)#2 (0) { }

And after, checking $this->_conn inside multiInserts() :
var_dump($this->_conn) = object(PDO)#2 (0) { } 

So, the #2 tells me it's the same object! (or am I mistaking?). It should be good. I still tried passing the $con object to my statement constructor as a 'copied' value instead of a reference, and problem remains.. changes nothing!

So, I'm very confused here as to what could be happening ? As additional info, my original $con object passed to the statement has the following PDO Attributes
 

$pdoAttribs = [
  PDO::ATTR_ERRMODE => PDO::ERRMODE_EXCEPTION,
  PDO::ATTR_DEFAULT_FETCH_MODE => PDO::FETCH_ASSOC,
  PDO::ATTR_EMULATE_PREPARES => false,
  PDO::ATTR_AUTOCOMMIT => false,
  PDO::ATTR_PERSISTENT => true
];

Also, I do not have the exact same versions of PHP / Lighttpd on my server as on my working PC

Server:
PHP 7.1.4 
Lighttpd/1.4.32

PC:
PHP  7.4.3
Lighttpd/1.4.55

But, it's complicated to get the same versions installed everywhere... HOWERVER, I did at least match the configurations for both lighttpd and php configs. 

Here's a link to my phpinfo() : http://corbeauperdu.ddns.net/phpinfo.php

And as for my lighttpd.conf :
 

# Special Configuration file for the DNS-325
# web servers are declared as virtual hosts

# include, relative to dirname of main config file
include "/etc/lighttpd/lighttpd.conf"

# adding extra modules
server.modules += ( "mod_redirect" )
server.modules += ( "mod_compress" ) # PRoy: enable compression module
server.modules += ( "mod_expire" )   # PRoy: used to set expiry time for cache on clients

# add support for new virtual host
$HTTP["host"] =~ "(^|\.)myhost.ddns.net$" {
  server.document-root = "/srv/WWW/html"
  server.errorlog = "/srv/WWW/logs/error.log"
  server.error-handler-404 = "/blog/index.php"
  accesslog.filename = "/srv/WWW/logs/access.log"

  # files to check for if / is requested
  index-file.names = ( "index.php", "index.html", "index.htm" )

  ###########
  # PRoy: compress content to DISK as to save CPU-cycles when requesting these file types (otherwise, CPU will compress it all everytime!)
  compress.allowed-encodings = ("gzip", "deflate")
  compress.filetype  = ("text/plain", "text/html", "text/javascript", "text/css", "text/xml" )
  compress.cache-dir = "/srv/WWW/tmp/wwwcache"

  ###########
  # PRoy: setup CACHING on clients
  # make clients think their content is already up-to-date
  # mod_expire will actually send a Cache-Control / Expires HTTP header on each item it's called upon (css, jpg, or etc) or called URL path
  #
  # Cache based on suffix
  $HTTP["url"] =~ "\.(jpg|gif|png|css|js|eot|svg|ttf|woff|woff2)$" {
    expire.url = ( "" => "access plus 13 months" )
  }
  # Cache based on path
  expire.url += ( "/calchom/css/"  => "access plus 13 months",
                 "/blog/videos/"  => "access plus 13 months",
  )

  # Anything else not caught by mod_expires will be cached on client with these directives
  setenv.add-response-header += (

    # cache for 1 month
    #"Cache-Control" => "public, max-age=2592000, must-revalidate, proxy-revalidate"
    # no cache at all!
    "Cache-Control" => "no-store, no-cache, must-revalidate, proxy-revalidate, max-age=0, post-check=0, pre-check=0"
  )

  ###########
  cgi.assign = (".cgi" => "")
  static-file.exclude-extensions = ( ".php", ".pl", ".fcgi", ".cgi" )
  flv-streaming.extensions = ( ".flv" )

  ##########
  # PHP Server configuration php-cgi vs php-fpm)
  # PHP-CGI (2020-02-02 : works good!)
  # Normally, PHP-FPM should be faster, but not on NAS (?)
  fastcgi.server = ( ".php" => ("localhost" => ("bin-path" => "/ffp/bin/php-cgi", "socket" => "/tmp/php-cgi.socket", "bin-environment" => ( "PHP_FCGI_CHILDREN" => "0" ), "max-procs" => "1") ) )

  ###########
  server.upload-dirs=( "/srv/WWW/tmp/wwwuploads" )
  server.max-request-size = 4194303
  server.network-backend="writev"

}

I know this is a long post, and a long shot.... Hopping to hear back from you folks.

Million thanks!!
Pat