Bug? Apache/php5-fpm returns 500 error over ssl, 200 over http when connection lost

[rocky10]

I'm having a really strange issue that only started when I changed my site to use SSL. Basically, If I disconnect from apache during a request originating from the server when using https it throws a 500 error, if I do it over http it just returns 200. I've tried this using the apache php module (not fpm) and it doesn't have the problem. It also doesn't happen if I access the server remotely. It only happens making the request on the server. There are multiple sites on the server so it is using SNI. The page loads fine if it is not interrupted.

To recreate:

Put this script on the server (it just sleeps for two seconds so you have a chance to interrupt it):
delay.php:

<?php
sleep(2);

Now, do a curl or wget request of that file on the server, and interrupt it before the two seconds by pressing Ctrl+C.

curl http://[site]/delay.php
^C

Check apache log:

"GET /delay.php HTTP/1.1" 200 163 "-" "curl/7.35.0"

Note the 200 response.

Now try with https (again interrupt before the two seconds by pressing Ctrl+C):

curl https://[site]/delay.php
^C

Check apache log again:

"GET /delay.php HTTP/1.1" 500 5064 "-" "curl/7.35.0"

*edit*
The error log contains:

AH00524: Handler for fastcgi-script returned invalid result code 32

Note it now responds with a 500 internal server error. If I don't interrupt the request it completes successfully and returns a 200 response.

I'm using the latest packages in Ubuntu Server 14.04.2 LTS:
Apache/2.4.7 (Ubuntu)
PHP 5.5.9-1ubuntu4.9 (fpm-fcgi) (built: Apr 17 2015 11:44:58)

Apache config for php5-fpm:

<IfModule mod_fastcgi.c>
AddHandler php5-fcgi .php
Action php5-fcgi /php5-fcgi
Alias /php5-fcgi /usr/lib/cgi-bin/php5-fcgi
FastCgiExternalServer /usr/lib/cgi-bin/php5-fcgi -idle-timeout 600 -flush -host 127.0.0.1:9000 -pass-header Authorization
</IfModule>

Anyone have any ideas?

I originally noticed this on a WordPress site because wp-cron was throwing seemingly random 500 errors after I upgraded to SSL. It turned out it was only happening when wp-cron uses curl with a very small timeout and the cron script takes longer than the timeout to run. Here is a wordpress forum thread about the issue: http://wordpress.org/support/topic/w...s-on-ssl-sites