Is this what we want as a default? This default results in much worse first-year, full-depth performance, and I need to remember to change it when writing about what I consider to be "baseline" performance. I'd propose updating "delay" to 0.
Relatedly, my impression is that the default plan is now depth-first, so I'd consider renaming rules.yaml to rules-breadth.yaml and renaming rules-depth.yaml to rules.yaml. Or alternatively renaming rules.yaml to rules-breadth.yaml and copying rules-depth.yaml to rules.yaml, so that the default out-of-the-box is closer to what people are now expecting.
The default fiberassign cadence is once a lunation, with a delay of 1. https://github.com/desihub/desisurvey/blob/055a7b9896aa6c56cd78d9afa245fc8a2fbc975b/py/desisurvey/data/config.yaml#L136 I believe this means that at the end of each lunation we make QSO decisions not for the most recent lunation's observations, but for the previous lunation's observations.
Is this what we want as a default? This default results in much worse first-year, full-depth performance, and I need to remember to change it when writing about what I consider to be "baseline" performance. I'd propose updating "delay" to 0.
Relatedly, my impression is that the default plan is now depth-first, so I'd consider renaming rules.yaml to rules-breadth.yaml and renaming rules-depth.yaml to rules.yaml. Or alternatively renaming rules.yaml to rules-breadth.yaml and copying rules-depth.yaml to rules.yaml, so that the default out-of-the-box is closer to what people are now expecting.