I’m joining the all-green meta goal experiment using this custom goal:
https://www.beeminder.com/narthur/buffer#data
It’s using the max aggregator with non-cumulative data points. I’m using a cron job to periodically post data points to it with the value being equal to the lowest lane value of all my goals (excluding /buffer itself) multiplied by each goal’s yaw.
The goal’s slope combined with non-cumulative data points means that I’ll need to increase my buffer over time until I reach the point where every goal is above the road.
Using lanes like this means the changes in required buffer are probably going to happen in sudden jumps instead of a gradual climb–I think I’d need to base the goal on days of buffer instead of lanes to avoid this. Not sure I made the right decision here, but I’ll give it some time before I decide to change it…
We’ll see how it goes… 
EDIT: The more I thought about it, the more using lanes just didn’t feel like the right choice. I’ve modified the script to use days of buffer instead of lane numbers, which should make the goal work much better since it will enforce a gradual change in my behavior instead of stair-stepped mega beemergencies.
Also, with this approach I should be able to schedule breaks by setting the rate to -1 (losing one day of buffer per day), which, after I’ve built up enough buffer in all my goals, could mean I’ve accidentally built myself the one-vacation-dialer-to-rule-them-all feature…