docs: initialize project wiki from production hardening session

12 pages covering architecture, findings, and metrics from the
rotatingMachine + machineGroupControl hardening work:

- Overview: node inventory, what works/doesn't, current scale
- Architecture: 3D pump curves, group optimization algorithm
- Findings: BEP-Gravitation proof (0.1% of optimum), NCog behavior,
  curve non-convexity, pump switching stability
- Metrics: test counts, power comparison table, performance numbers
- Knowledge graph: structured YAML with all data points and provenance
- Session log: 2026-04-07 production hardening
- Tools: query.py, search.sh, lint.sh

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

wiki/sessions/2026-04-07-production-hardening.md

@@ -0,0 +1,46 @@
+---
+title: "Session: Production Hardening rotatingMachine + machineGroupControl"
+created: 2026-04-07
+updated: 2026-04-07
+status: proven
+tags: [session, rotatingMachine, machineGroupControl, testing]
+---
+
+# 2026-04-07 — Production Hardening
+
+## Scope
+Full code review and hardening of rotatingMachine and machineGroupControl nodes for production readiness.
+
+## Key Discoveries
+
+1. **Efficiency rounding destroyed NCog/BEP** — `Math.round(Q/P * 100) / 100` in canonical units (m3/s and W) produces ratios ~1e-6 that all round to 0. All NCog, cog, and BEP calculations were non-functional. Fixed by removing rounding.
+
+2. **flowmovement unit mismatch** — machineGroupControl computed flow in canonical (m3/s) and sent it directly to rotatingMachine which expected output units (m3/h). Every pump stayed at minimum flow. Fixed with `_canonicalToOutputFlow()`.
+
+3. **emergencyStop case mismatch** — `"emergencyStop"` vs config key `"emergencystop"`. Emergency stop never worked. Fixed to lowercase.
+
+4. **Curve data anomalies** — 3 flow values leaked into power columns in hidrostal-H05K-S03R.json at pressures 1600, 3200, 3300 mbar. Fixed with interpolated values.
+
+5. **C5 pump non-convexity** — 5 data points per pressure level produces non-convex spline. The marginal-cost refinement loop closes the gap to brute-force optimum from 2.1% to 0.1%.
+
+## Changes Made
+
+### rotatingMachine (3 files, 7 test files)
+- Async input handler, null guards, listener cleanup, tick loop race fix
+- showCoG() implementation, efficiency variant fix, curve anomaly detection
+- 43 new tests (76 total)
+
+### machineGroupControl (1 file, 2 test files)
+- `_canonicalToOutputFlow()` on all flowmovement calls
+- Absolute scaling bug, empty Qd block, empty-machines guards
+- Marginal-cost refinement loop in BEP-Gravitation
+- Missing flowmovement after startup in equalFlowControl
+
+### generalFunctions (1 file)
+- 3 curve data fixes in hidrostal-H05K-S03R.json
+
+## Verification
+- 90 tests passing across both nodes
+- machineGroupControl within 0.1% of brute-force global optimum (1000-step search)
+- Pump switching stable: 1-2 transitions across full demand range, no hysteresis
+- Optimization cost: 0.03-0.15ms per call (0.015% of tick budget)