Thanks for your comments. Regarding my earlier test of various audio frequencies, those numbers were rounded estimates with no attempt to be on 500 Hz or multiples. I just wanted to try to vary the audio frequency to eliminate that as a part of the puzzle.
I have clamp-on chokes on both ends of the USB cable between PC and K3-S -- each with two turns in ferrite.
Regarding DC current drain, I have an ASTRON RS-35M with meters. In CW mode at 100 watts, the meter shows about 18 Amps (not an accurate meter). When transmitting a CQ message in WSJT-X, the current varies between 18 and 14 amps, approximately.
All measurements are with a dummy load. Also, I removed unnecessary cables -- Key, Paddle, earphones, Mic.
Also, reducing the power to 25 watts or so still has significant fluctuation of DC current and RF out as viewed on P3 with TX monitor. Further reduction to 12 watts (hear relay click for 100 watt amplifier) is difficult to judge because current drain deflection of meter is too small and P3 display is also a small variance in amplitude.
Thanks again for your suggestions.