This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
The last pad collisions came when the whole queue moved one waiting spot closer at once, with craft swapping heights. A gust could stall one craft mid-move.
What was tested
Six queue rules were tried. Then one-at-a-time moves: a craft leaves its spot only once the craft ahead has physically left the next one.
What was found then
All six queue rules either jammed the incoming wave or put swapped neighbours at one height. One-at-a-time moves worked. In short turbulent test runs, collisions fell from 6–7 to 0–2 with the same number landed, for about 5 s of waiting per move. The full test set (every scenario family and variant, about 400 cases) went from 33 collisions to 21. Every other family was unchanged, and 1 % fewer craft landed in the pad families.
What changed
One-at-a-time moves in the pad approach lanes.
Corrections retained in the record
Earlier wording: Eight queue rules were tried and rejected.
Record correction: Six queue rules were tried (plus a wider swap hysteresis, which tied).
README.md:284
Numbers and source references
- Six
- queue rules tried and rejectedREADME.md:284
- 6–7 collisions to 0–2
- turbulence-3 probe runsREADME.md:328
- 5 s
- wait per hopREADME.md (Stress campaign) (Sequential hops)
- 33 to 21
- full-matrix collisionsREADME.md (Stress campaign) (Sequential hops); git 42d0f67
- 1 %
- fewer craft landed in the pad familiesREADME.md:332