many years ago we had a ...

many years ago we had a similar issue at a large brand plant. What we did then was to take an easier path and placed a conductivity monitor at the influent to the bug plant and data logged it (actually we used a chart recorder). This gave us the time periods of dumps. We then when backwards into the streams that fed the main. The majority of the spikes were related to cleaning operations and nearly everyone was a high pH dump as well. 

The study provided an unexpected benefit. The bottle or can runs were scheduled by cases.  If 200 cases were scheduled, 200 cases were filled. Whatever else left, was dumped to waste. The process was changed to where the line was run until empty. Partial cases were sent to the lobby venting machines and given to visitors. Remember the opening words, Many years ago.

I'd still use something simple to 1st identify the major abusers and then see what is in those once isolated.