Ok, so I looked at this for a good chunk of my evening… The issue was caused by whatever method they used to “restore” the site from backup. You ended up with some sort of frankenstein cross between 8.2.10 and 8.3. In 8.2.10 I had a class named cnLog() this class was pretty simple and it main purpose is used to debug image issues. This logs created by this class are stateless, meaning the logs are only available for a single page load. Well, in 8.3 the cnLog class was changed to a stateful log, meaning the logs are saved in the database so they can be reviewed and managed at any point in time. The old cnLog is still available in 8.3, it was renamed to cnLog_Stateless so it still could be used for image debugging.
Because of this frankenstein hybrid you ended up with the cnLog was the new stateful version. And this was bad, very bad… did I say bad? all those logs were being saved to the db which was basically bringing your server to its knees.
Now you have tens of thousands (hopefully not more, but possible) of orphaned log data cluttering up your db. We’ll come back to this…
I was able to get your directory back up by deleting Connections and reinstalling a good copy of 8.2.10.
No, those logs, all those completely useless/wasteful logs… those can be gotten rid of, in time. There is no way to purge them without bring your server to its knees again. And the only way to start removing them it to get you updated to 8.3.1. So, please update. I believe the issue you had initially is fixed. If you have trouble with the update, please do not contact the host to fix it, please contact me.
After you update, let me know, I’ll add a little code snippet that will purge those logs a little bit at a time once an hour.