TPA3255: Output nearly-dead; advice on debugging steps

The design uses the dual BTL  (M1 = 0, M2 = 0), and it is a revision ("second generation") of a previous design that was working ok.

This is the partial schematic  (most of the decoupling network not shown;  but it is essentially identical to what's suggested in the datasheet):

(notice, RT7 is the OC resistor, I set it to 22k;  RT8 is the FREQ_ADJ resistor, set to 10k for AM2 — 600kHz — PWM frequency)

I'm applying 30V to all PVDD voltages, and a 12V wall adapter (producing 12.1V as measured at pin 2 of the TPA3255).

I confirmed (via test points) that M1 and M2 are both set to 0.

The symptoms:

The output activates and produces exactly 10 pulses (i.e., if I look at one of the output pins; e.g., OUT_A) before it shuts off.  The frequency is dead-on:  6 pulses in exactly 10 μs.   At the time that the output dies, FAULT is asserted, just for a little while.  The process repeats maybe every 100ms or 200ms?  (I can see the FAULT LED blinking more or less fast)

The output does not seem shorted.  I check with an ohmmeter, and I get: between either one of the outputs (OUT_A, B, C, or D) and GND, approximately 250 kΩ;  between OUT_A and OUT_B, approx. 700 kΩ;  same thing between OUT_C and OUT_D.

At first, the output had no load whatsoever  (in the design, the LC output filter is on a separate board).  I connected a light resistive load just in case  (200 Ω between OUT_A and OUT_B, 200 Ω between OUT_C and OUT_D).  Did not change anything.

Additional "debugging information/hints" I measured:

• Voltage at OSC_ADJ pin:  1 V
• Voltage at FREQ_ADJ pin:  approx. 0.8 V
• Voltage at C_START pin:  0 V
  
  
• Pin 2:  12.1 V
• Pin 1:  10.55 V
• Pin 22:  11.5 V
  
  
• Pin 11:  3.31 V
• Pin 14:  7.82V

Any ideas of what else I should try / measure?  (or in any case, anyone recognizes the symptoms and can guess where the culprit may be?)

Thanks,
Carlos
--