[JKRT]

Takes the relaxed validation settings (atol/reltol above 0.01/3e-3, a stopTime short of the experiment's) off the relaxed entries of models that validated: 113 for MSL 4.1.0, 123 for 3.2.3. Solvers, observed filters, mappings and reference files are unchanged; a stopTime stays only where the reference ends before the experiment (ChuaCircuit 4.1.0 goes to its reference's end, 3000 s).

Checked by running each relaxed model in OM.jl through the harness at the default rule over the full horizon, and against omc on a sample:

• Strict passes (95 for 4.1.0, 105 for 3.2.3): the relaxations were stale, the count does not change.
• Strict failures, 18 per version: they now count as failing; each issue notes the finding.
  • Both versions: Electrical.Digital BUF3S, DFFREG, DFFREGL, DFFREGSRH, DFFREGSRL, HalfAdder, INV3S, NRXFER, NXFER, WiredX (OM.jl's logic outputs differ, omc matches the reference); OpAmps.SchmittTrigger; MultiBody PointGravity; FluidHeatFlow OneMass (3-5 K off), PumpDropOut and ParallelPumpDropOut (Unstable after the drop-out).
  • Disputed (omc also misses the reference): OpAmps.Multivibrator, MultiBody DoublePendulumInitTip, and FluxTubes ForceCurrentBehaviour (4.1.0) or ChuaCircuit's 50000 s chaotic reference (3.2.3).

Expected from the next coverage runs: MSL 4.1.0 about 482/534 validate (was 500), 3.2.3 about 370/425 (was 388).

🤖 Generated with Claude Code