One tool I don't have in my toolkit - a sniffometer. Nothing smelled while
in use or close handling of this power supply. But on removing the electros
, that distinctive fishy smell of leaked elecrolyte, with soldering iron
temperature on the pins.
Tiny space available to fit some capacitance so they used 2x squat format
22uF,35V in parallel. One leg pulled out of one cap on desoldering and the
other a telltale drop of brown gloop immediately under the cap but not
enough to be visible over the pcb.
Both bright green nichicon 105 deg C. Found space on the hot side for
replacement and powered up long enough at about 3KHz instead of 40KHz to
give full V on all o/p ( unloaded) with very audible 3KHz coming off the
transformer. As , presumably, the pair of switching power fets were
operating at least partially linearly at an undesigned 3 KHz they heated up
quite quickly. 15 V over that sensing zener . Removed the 27nF and now
operating correctly with substantial test loading on all rails.

One mystery remains , the original tick-tick was about 1 a second, on adding
the 27nF to the 555 timing capacitor the tick-tick rate increased to about
10 a second. I've not traced out the path from the "goodness of oscillation"
circuit to the 555 but what sort of intermediary cct would lead to an
increase in that tick-tick rate ?