This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
The previous stage never measured how long the emergency craft itself spends inside ordinary craft's safety bubbles, its last safety blind spot.
What was tested
A record-only counter timed every bubble entry on the previous stage's 12 test cases, comparing the default emergency craft (others give way to it) with a plain one. The one worse-looking case was flown again on fresh layouts.
What was found then
On 11 of 12 cases the two spent similar time deep inside bubbles. The exception, at 2× speed with a 1 s link delay, rested on one layout: 14.2 against 6.8 s per hour for the default. On fresh layouts it did not repeat. Neither comparison passed the threshold we fixed in advance for a real difference. That is not proof they are equal, nor a safety clearance: ordinary craft were still lost in large numbers.
What we learned later
Reviewers asked that 'alike on all twelve cases' be read precisely: eleven cases alike, plus one flown again on fresh layouts, with the first, worse-looking result kept on record. It was a check against a threshold, not proof that the two are equal.
research/reviews/researcher-c253.md
What changed
Nothing. A recording-only counter was added; runs are identical with it on or off.
Corrections retained in the record
Earlier wording: The episodes are short: 90 % end within 1.3–9.5 s, and the longest was 14 s. That is about the same as ordinary craft.
Record correction: The emergency craft's 90th-percentile episode ranged 1.3–9.45 s across cases; the ordinary fleet's was 3.9–4.6 s per run (maximum 8.9–19 s). The record calls these descriptive and adds that short episodes do not by themselves show the exposure is harmless; 'about the same as ordinary craft' is not in the record.
research/design/prio_breach_duration.md:55 (Reading, episode durations)
Earlier wording: ordinary craft are still lost in dozens, up to about 150 per run pair on the calm radio cases
Record correction: 140–159 ordinary craft lost on the calm 1 s cases, each figure pooled over the two seeds of one arm.
research/design/prio_breach_duration.md:55 (What this licenses)
Numbers and source references
- 48
- runs in the first measurement (12 cases × 2 craft × 2 seeds)research/design/prio_breach_duration.md:55
- 11 of 12
- cases where the two emergency craft were alikeresearch/design/prio_breach_duration.md:55
- 14.2 vs 6.8 s per hour
- time deep in the bubble, default vs plain, on the one worse case (2 seeds)research/design/prio_breach_duration.md:55
- 12
- fresh-seed runs of the worse case (6 seeds × 2 craft)research/design/prio_breach_fresh.md:43
- 5.9 vs 8.8 s per hour
- time deep in the bubble, default vs plain, on 6 fresh seedsresearch/design/prio_breach_fresh.md:43
- 416 and 412
- ordinary craft lost over six half-hour runs for each emergency-craft version, fresh layoutsresearch/design/prio_breach_fresh.md:43
- 3–3
- layout-by-layout split on the fresh layoutsresearch/design/prio_breach_fresh.md:43
- 2×
- emergency-craft speed on the exception caseresearch/design/prio_breach_duration.md:55
- 1 s
- link delay on the exception caseresearch/design/prio_breach_duration.md:55
- 6
- fresh layouts for the re-flightresearch/design/prio_breach_fresh.md (Spec, Rows)