Model 1.1.0, November 15, 2026. Baseline 1.0.0: 14/14 rerun before modifications.
Candidate command: node --test tests/*.test.cjs

✔ bounded search enumerates 567 distinct phase policies from identical initial conditions (0.685167ms)
✔ recommended results independently reproduce their exact shared-engine metrics (9.257791ms)
✔ constraint results distinguish survival from planning reserve requirements (1.767667ms)
✔ each returned frontier point is feasible and no candidate dominates it (0.335583ms)
✔ recommendations expose competing scientific work, greenhouse care and reserves (0.056541ms)
✔ science recommendation maximizes science within feasible search grid (0.095375ms)
✔ sensitivity executes fixed decisions under five explicit configurations (6.366709ms)
✔ robust recommendation satisfies every tested condition without retuning (8.706458ms)
✔ policy and configuration failures are explicit, not hardcoded wins (1.049792ms)
✔ every frontier policy produces an exact replayable flight record (18.375083ms)
✔ valid but severely underpowered conditions yield no invented feasible recommendation (96.9935ms)
✔ NASA surface rates give 0.07 kg oxygen and 0.266666667 kg water per crew-hour (1.796542ms)
✔ safe decisions complete the mission with science and reserves (3.23375ms)
✔ ignoring the array fault causes a reproducible power failure (0.626375ms)
✔ repair without shelter fails the exposure risk constraint (1.479292ms)
✔ disabling oxygen production exhausts stored oxygen (1.148125ms)
✔ disabling recovery exhausts the water reserve (1.906458ms)
✔ same decisions yield identical state including logs (9.301791ms)
✔ repair takes two powered intervals and changes the following interval (0.372375ms)
✔ shelter pauses science and adds load while reducing risk (0.704292ms)
✔ resource and energy accounting includes recovery reject and electrolysis hydrogen mass (0.127458ms)
✔ oxygen production stops at storage capacity and spares feedwater (0.079583ms)
✔ terminal states are immutable and input state is never mutated (2.269125ms)
✔ invalid controls fail loudly (0.210792ms)
✔ long mission policy sweeps preserve bounds and finite state (16.551958ms)
✔ dry food inventory follows NASA Table 2 dry-mass planning rate (0.14375ms)
✔ greenhouse competes for power and water without creating instant food (0.135958ms)
✔ a six-hour simulated outage buffers data and enabled radio later drains it (2.286125ms)
✔ disabled radio retains bounded data and reports overflow rather than inventing delivery (2.195333ms)
✔ each interval conserves generated, delivered, buffered and dropped data (2.081417ms)
✔ greenhouse water and curtailment appear explicitly in physical ledgers (0.094084ms)
✔ uncertainty changes supply, consumption and repair duration through the shared equations (0.651542ms)
✔ configuration rejects unknown, nonfinite and out-of-range conditions (0.142375ms)
✔ impossible states fail loudly before prediction or stepping (0.438042ms)
✔ configuration and full decisions replay exactly under uncertainty (4.54725ms)
✔ unknown and malformed decisions cannot silently alter the model (0.131375ms)
✔ all 64 constant system combinations retain finite bounded states (67.665917ms)
✔ scarce water records actual crew use and unmet demand without negative inventories (0.145084ms)
✔ scarce oxygen and food distinguish served inventory from unmet demand (0.128417ms)
✔ wastewater overflow and treatment reject are explicit ledger outputs (0.134625ms)
✔ tracked water/oxygen conversion and wastewater accounting close for every completed interval (2.11525ms)
ℹ tests 41
ℹ suites 0
ℹ pass 41
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 1226.708958

Real Chrome/Playwright browser checks:
PASS: real manual recovery, mode/state invariance, glossary, keyboard map, repair, food/data outputs, exact exported replay, computed policy comparison and export, five sensitivity cases, applied robust policy, what-if restart, judge playback outcomes, mobile bounds, offline reload and direct-file planner; zero page errors.

These verify software behavior, not flight qualification, classroom outcomes or accessibility certification.