Correction to #3534: Chris's 0.45 ms tail was taken with a wet cell (4 ppm) and probes on the plates, so it is a drive tail, not stored charge; the step-charge plateau match is withdrawn

Chris Freedom, Discord #neuralstan20-testing 2026-11-08. He confirms that for shot (b) both probe tips were on the exciters, both ground clips were on the secondary 0 V (CT) node, and the cell was WET with 4 ppm water. Consequences: (1) Probe loading is ruled out, because 10 MΩ (yellow, 10x) and 100 MΩ (blue, 100x) are negligible against roughly 12 kΩ of water. (2) Reading yellow ÷10 (the 10x probe was on the x100 setting, #3532): plate+ is about +170 V at the spike and about +60 V at the start of the tail; plate− is about −70 V and −37 V. That leaves roughly 100 V across the cell for hundreds of µs. The water's own hold time is ε/σ ≈ 1.2 µs, so this is not stored charge. About 8 mA is flowing through the water throughout, which means the transformer is still driving. (3) Best-fit hypothesis, unconfirmed: the MOSFET is still on (forward mode). Each 380T half gives 10×Vin, and Vin = 4 V (#3528) gives 40 V, close to the −37 V blue tail. Primary L ≈ 91.5 mH × (38/380)² ≈ 0.92 mH. The primary voltage then decays as e^(−tR/L) while magnetizing current climbs through R_DS(on) plus wiring plus supply resistance; τ = 0.45 ms needs R ≈ 2 Ω. (4) The step-charge match in #3534 (V∞ ≈ 1.2× by pulse 4) assumed the cell held charge for τ ≈ 0.45 ms. That assumption is withdrawn. Discriminating test requested: gate on CH1 and a plate on CH2 at 100 µs/div; then double Vin (the tail amplitude should double while τ stays roughly the same).