This variant , they've sorted out the bad thermal air path "design" for
the main heatsink, but introduced another.
Zobel network involving 2x 4.7R 5W in parallel litterally back to back
and also hemmed in, touching, by the caps, relay and the copious white
gunk.
A pathetic looking flat baffle plate that presumably was someone's idea
that the inlet air would then magically divert over the tops of these
Rs, unlikely.
So instead of 10W rating , presumably effectively derated to about 3W.
Etiology looks as though PbF failure of one R solder joint (burnt pcb
hole in the pcb around the pin ) leading to that R actually cracking up
and going o/c and eventually the other one going high, and along the way
the heat cooked one of the paralleled caps to short circuit . Also
cooked and melted relay case , so the contact is permanently closed,
explains the switch on and off thumps/chirps , over a few weeks, that
did not used to be there.