What the solver did each tick, and whether mass was kept: the solver tab.
A run that converges, plots smoothly and violates no assertion is not evidence
(docs/PHYSICS-BENCHMARK.md opens with that). This is the other half: the
numbers a CFD code shows beside every answer, so a person can see when one is
not to be trusted.
Per tick, a :class:`SolverRecord`:
* **Newton.** How many coupling steps the tick took, the network iterations
they cost, the worst scaled residual and whether every solve converged.
* **Continuity.** The worst net mass imbalance at any free node of the network
[kg/s] -- the solve's own conservation, which a converged solve can still
violate if its tolerance is loose.
* **Chamber closure.** How far apart the network's flow and the engine's
chamber pressure were left [psi].
* **Global mass balance.** Everything held in the vessels, against everything
that crossed the stand's boundary: the engine, vents, a drawn supply, and the
built-in loads and charges. Cumulative, so a leak of micrograms a step shows
as a line that drifts. Zero is the only right answer.
* **Integrator guards.** What the vessels' own floors and clamps changed beyond
what their rates said -- mass [kg] and ullage energy [J], cumulative. Each
guard is physics (a vent cannot pull a tank below atmosphere), but each one
changes the inventory, and seeing how much is the point.
Read-only: nothing here changes a step.