After 2 hard days of dead ends, bodges, & blown traces, the problem was solved. The always dependable LM317 was creating noise in the muted state. The LM317 was always considered a very quiet voltage regulator compared to the LM7805, reliably powering MEMS sensors & audio preamps for years. This was the 1st time it proved too noisy & there was nothing in the probing to reveal it. It was just a matter of eliminating every other change since the through hole board.
A 10k 100uF RC filter got the noise back to what it was on the through hole circuit, making this the 1st SMD iteration which actually worked. The mane problem in previous SMD boards was slightly tweeking the connector causing an overload feedback oscillation. That hasn't happened with board #3 so far & the noise was the only other issue.
The mane question has been adding an LED for mute mode, but the only way to know if the LED indicates muted or on is to listen on the phones, so it's quite useless for anything but bling. Helas, this board is already nothing but bling. The giant knobs & switches are completely superfluous & it would be better to have the power LED next to the power switch.
It would be a matter of whacking in a 2nd mute transistor to turn on the LED when muted. It would be a lot harder to turn off the LED when muted. Back when we commuted to offices with land lines, the old polycom phones lit a red LED when they were muted even though no-one knew what it meant in practice.
Also, upgrading the LEDs to giant size would be another superfluous bling on top of the bling.