So when I did a test run with the amendments suggested (sheets of paper instead of glass panes!), there was nowhere near enough space. So I've come up with this instead.
The idea of the first pane on the left is just to keep the surface scum where the 9004 can pull it in, rather than going straight back to the return pump
First and second chamber to stay at the same (constant) level. Final chamber allowed to fluctuate for ATO.
Does the drop in the baffle look long enough to get rid of the bubbles?