I said to Claude: The leads to the diode at the cut point are fine. So we are saying what I measured again and again can't have been true. I am seriously struggling with this. I literally went from one amp to the other in the same session. 0V on the busted amp, 130V on the working one. Clear as day.

Claude said: Let's step back, because I think there's a way to reconcile everything you've measured without doubting yourself — and it's something we touched on very early today but never fully resolved.
A shorted diode would produce exactly this pattern. Here's why: if D001 or D002 has failed shorted (not open), it effectively places a dead short directly across that half of the transformer's secondary winding whenever the circuit is complete and powered. Under a genuine short, voltage at that point collapses toward 0V (V=IR, and R≈0 across a short) — while current shoots up massively, which is exactly what would blow R003 on the primary side. This unifies both of your major findings into one single fault: the 0V reading was real, and it explains the original symptom that started this whole investigation.
It also explains why the winding tested perfectly healthy once removed: a shorted diode isn't a transformer fault at all — the winding was never the problem. It only showed 0V because something downstream was short-circuiting it under load, not because the winding itself was open. Out of circuit, with nothing loading it, of course it measured fine.
Why this wasn't caught earlier: you did test these diodes for shorts, but the test relied on comparing against the working unit's continuity pattern — and the working unit's diodes showed the same "beeps across all three legs" behavior, which you and I concluded was normal packaging (case tied to one terminal internally). If that packaging conclusion was right, a genuinely shorted diode on the faulty unit might have looked identical to healthy-but-normally-tied-together packaging, and the continuity test simply couldn't tell them apart. That's a real limitation of that test, not a mistake either of us made carelessly.
How to test this properly now: with the diode fully isolated (not in-circuit, ideally desoldered or at minimum both leads lifted), measure it in diode-test mode, not continuity/beep mode — you're looking for a proper forward voltage drop in one direction and a clean open in the other. A stable low-resistance reading in both directions on an isolated diode (not just a case-related beep) would be unambiguous confirmation of a short — no packaging ambiguity possible once it's truly isolated.
This feels like the missing piece — worth testing before anything else.