This is an earlier experiment. Read the amendments below before treating its findings as current.

The question

Why are crossing-lane craft within reach 200–300 m before the crossing, and do they arrive before a craft turns away?

What was tested

No new runs: for each crossing passage, which came first, neighbour or turn; and each craft's time far off its lane.

What was found then

Where both happened, the neighbour came first in 75–88 % of passages, about 20 s before the turn, even with no crossing craft near at the start. Early neighbours were crossing-lane craft 258–307 m off their own lane, at the edge of its spread. With heading copying on, 10 % of ideal-craft (instant-response) positions are over 219 m off lane; with it off, 99 % are within about 89 m, and no early neighbour appeared in 23,500 passages. Most ideal craft (84–88 %) stray for about a minute, and nearly all wing craft stray. Whether craft react or pass the disturbance on stayed open.

What we learned later

LATER / CYCLE 100

Counting all passages, the neighbour came first in about 7 in 10 (68–73 %), and the 'one in twenty' share where the turn came first is not a hard minimum.

research/envelope.md:5229

LATER / CYCLE 101

Withdrawn: 'the wing turns before any neighbour arrives' (a mistake in the search window). There is no sideways wall in the code, so the 89 m is an observed spread, not a limit.

research/envelope.md:5152

LATER / CYCLE 102

Withdrawn: 'excursions mostly start in mid-flight'. About half follow the craft being put back at the lane start after reaching its end. Counted across those restarts, an excursion lasts 70–75 s (cycle 103).

research/envelope.md:4991; research/envelope.md:4940

What changed

Nothing.

Numbers and source references
75–88 %
passages with both events where the neighbour came firstresearch/envelope.md:5229
20 s
typical lead of the neighbour over the turn (17–22 s)research/envelope.md:5229
258–307 m
median off-lane distance of early crossing-lane neighboursresearch/envelope.md:5152
219 m
90th percentile off-lane distance, ideal craft, copying onresearch/envelope.md:5152
89 m
99th percentile off-lane distance with copying offresearch/envelope.md:5152
23,500
passages with copying off and no early neighbourresearch/envelope.md:5152
84–88 %
ideal craft that spend time more than 200 m off laneresearch/envelope.md:5094
about a minute
typical excursion before returning (fragment median)research/envelope.md:5094
98–100 % / 28–49 %
wing craft visiting the far-off-lane state / their median share of time thereresearch/envelope.md:5094
Original structured record ↗