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

The question

Speed had shown no safety limit from 4 to 30 m/s, but only for the ideal craft (instant response).

What was tested

Calm, 600 craft, half-hour runs, 3 layouts each: wing and ideal craft at 6–30 m/s; multirotor, eVTOL and a bare slow-responding test craft at 8 and 16 m/s.

What was found then

The wing lost about 10 craft per run at 8 m/s and 51 at 16: 5× the losses for twice the speed. The ideal craft lost none. Every slow type paid, 3.9× to 10.2× from 8 to 16 m/s. At 16 m/s losses ranked by lag (21 per run at 0.5 s, 55 at 1 s); the wing roughly matched the bare test craft. At 22 and 30 m/s the wing lost 95 and 189 per run, and the ideal craft had its first speed losses: 4 craft in one run of three at each. Lag was named the likely reason.

What we learned later

LATER / CYCLE 119

Narrowed: speed shows a steady rise, not a line between safe and unsafe; why lag kills was restated as an idea still to be proven.

research/envelope.md:4332

LATER / CYCLE 120

Narrowed: lag as the reason is a hypothesis, because the types also differ in acceleration. The wing matching the bare test craft (ratio 0.94) means 'within a set range', not proof they behave the same.

research/envelope.md:4280

LATER / CYCLE 121

The ideal craft's losing runs held two collisions each, not one.

research/envelope.md:4237

LATER / CYCLE 148

Stage 2 later adopted a steering rule that allows for the delay; it halved losses, but losses still grow with speed.

research/design/phase1_close.md; research/design/tau_stage_close.md

What changed

Nothing.

Numbers and source references
600 / 1,800 s
starting craft and run lengthresearch/envelope.md:4332
24 / 18 / 12
runs in the three speed stepsresearch/envelope.md:4332; research/envelope.md:4280; research/envelope.md:4237
12 / 10 / 27 / 51
wing craft lost per run at 6 / 8 / 12 / 16 m/sresearch/envelope.md:4332
5×
wing losses at 16 vs 8 m/s (5.13)research/envelope.md:4332
3.9× / 4.9× / 10.2×
loss growth 8 → 16 m/s: multirotor / eVTOL / bare modelresearch/envelope.md:4280
21 / 26 / 55
losses per run at 16 m/s for lag 0.5 / 0.75 / 1 s (20.7 / 26 / 54.7)research/envelope.md:4280
0.94
wing vs bare model loss ratio at 16 m/sresearch/envelope.md:4280
95 / 189
wing craft lost per run at 22 / 30 m/sresearch/envelope.md:4237
4
ideal craft lost in one of three runs at each of 22 and 30 m/sresearch/envelope.md:4237
Original structured record ↗