Noticed a significant droop on low frequency square waves, say 10-100Hz. Changed C1 to a 1uF 50V mylar.
Next, if thought I have a look at trying to get rid of the artifacts (noise spikes) which I see on the 1kHz test signal at a timebase setting of 0.5ms. With another oscilloscope and probing around the various supplies I noticed that there are some quite high voltages across L2. I wondered if replacing the +9V to -9V discrete SMPS with a Maxim ICL7662 CMOS voltage converter would make any difference.
So, to save someone else the trouble it made no difference!