This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
One collision remained in still air: the pad where all craft arrive at once, in dense fog.
What was tested
We read the trace, then changed the rule that lets queued craft swap order, so that a swap never gives two nearby queue-mates each other's heights.
What was found then
Three craft that joined close together swapped order twice in 1.5 s. Each swap exchanged heights, sending one craft climbing into another descending 13 m away. Swaps are now refused at close range. Instead, an out-of-order pair two spots apart trades spots at the same height. The fog case cleared: closest approach rose from 0.81 to 1.06 of the required spacing (below 1 means a safety bubble was entered). The full test set (every scenario family and variant, about 400 cases) went from 1 collision to 0. Every non-pad result and the work total stayed exactly the same.
What changed
A height-aware swap in the pad queue, with same-height spot trades.
Numbers and source references
- 1.5 s
- time between the two swapsREADME.md:449
- 13 m
- horizontal gap of the colliding pairREADME.md:449
- 0.81 → 1.06
- closest approach in the fog row, in standardsREADME.md:449
- 1 → 0
- full-matrix collisions (run 18)README.md:449; git 73611d0