Computed the observation ceiling, which should have been done several iterations
ago instead of being approached one run at a time. A turn's shell call is capped
at 595 s; boot costs about 220 s, of which roughly 190 s is a title movie the
launch script waits out because tapping breaks the title; probe startup is about
25 s. That leaves roughly 350 s of observation, and at the measured 55 % of
real-time that is about 193 game-seconds.
Against the phase-1 route schedule: t=90 needs 164 s of wall-clock and t=120
needs 218 s, both already observed with no arrival. t=170 needs 309 s and is
reachable but only on a run that does not freeze. t=210 needs 382 s and t=240
needs 436 s, and neither fits in a turn at all.
So the tail of Stage 02's phase-1 schedule cannot be tested under this harness
however many runs are attempted. Recorded as a blocker rather than improvised
around. Two things would unblock it and neither is mine to decide: a safe way to
skip the title movie, which is more than half the budget and would roughly
double the window to about 297 game-seconds covering every entry, or a longer
shell timeout if the harness permits one.
This run is discarded: frozen at t=90 with 15 stalled samples and 3 losses,
correctly flagged and not interpreted. The freeze tally is now three clean of
five, lower than the three-of-four quoted last iteration.
What stands is unchanged: zero confirmed arrivals over 300 s of verified-live
flight, about 165 game-seconds, covering the t=90 and t=120 entries. That is the
strongest statement this harness can currently support.
324cb7f85a