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

The question

Six times the losses for three times the craft could mean more meetings (two craft inside each other's bubble) or deadlier ones. Shallow-angle meetings were a new lead on why delay kills.

What was tested

No new flights: the density and delay runs were re-read, then the link study was closed.

What was found then

With perfect information about 0.3 % of meetings ended in a loss at every density; the extra losses came from almost seven times more meetings. A 1 s delay made each meeting deadlier, 2.6× at 300 craft and up to 4.6× at 900. Longer delay raised the fatal share more in shallow meetings than in crossings, but not by the margin fixed in advance, so it stays an open lead. Closing summary: the link costs 2.7–5.6× more craft than perfect information, delay drives it, and every tested setting lost craft.

What we learned later

LATER / CYCLE 263

The recount had claimed every lost craft died in a collision between two ordinary craft. Slightly wrong: 1–6 per setting collided with the emergency craft. The counts now add up exactly.

research/design/s5_ending_share.md:39 (c263 note)

LATER / CYCLE 265

The pair-by-pair search, including the shallow-angle lead, was closed without finding why delay kills.

research/design/s5_angle_close.md:95 (Open)

What changed

Nothing.

Numbers and source references
about 0.3 %
fatal share of meetings with perfect information at 300 / 600 / 900 craft (0.26–0.37 %)research/design/s5_ending_share.md:39
almost seven times
more meetings per craft-hour from 300 to 900 craft (6.9×)research/design/s5_ending_share.md:39 (Density sentence)
2.6× / 4.4× / 4.6×
how much deadlier each meeting is with a 1 s delay at 300 / 600 / 900 craftresearch/design/s5_ending_share.md:39
99.5 %
minimum share of meetings that finish in any settingresearch/design/s5_ending_share.md:39
1.42 vs 1.5
shallow rise ÷ crossing rise (0.5 s → 2 s) against the preset thresholdresearch/design/s5_angle_close.md:54
2.7–5.6×
link cost vs perfect information across calm, wind and densitiesresearch/design/s5_angle_close.md:95 (Established 1)
3.1× / 2.2×
rise in fatal share from 0.5 s to 2 s, shallow meetings / crossings (3.08 / 2.17)research/design/s5_angle_close.md:54
Original structured record ↗