Royer oscillator not starting up?

[Mark]

I'm having a problem with the attached Royer circuit, 125VDC input, unregulated 24VAC output. 
The output then goes to a rectifier and 15V regulator.  Normal current draw is 0.06A from 125VDC input. 

[ Specified attachment is not available ]

It seems that when powered up from a slow ramping source there is no problem (e.g. powering up a 125VDC power supply from the mains), but when using a fast ramping source this circuit ocasionally blows the 0.5A fuse (e.g. switching 125VDC to the input). 

I had a hunch that the oscillator wasn't starting up properly, so I've tested this circuit thousands of times and only once managed to capture this:

pink trace shows 125V input,
blue trace is the current input,
Yellow and green are Q1, Q2 collector voltages. 




It looks like there is a delay in the oscillation starting, resulting in a current spike, then the oscillation starts but by that time the fuse is on its way out and the fuse blows after  37ms. 

Would you agree that it looks like the oscillator is not starting immediately and that is causing the higher input current? 
Has anyone seen this type of problem with Royer oscillators before and how would you suggest I fix the issue? 

For reference, this is what normally happens, the oscillations start immediately and the input current drops quickly to the 60mA level: 

[mag_therm]

Firstly, it is notable that the 2n3439 are able to clear the 0.5 A fuse without one failing.

In the old days the circuit sometimes had a starting resistor or RC  on one base so that transistor would always start first.
Another way is like your circuit, where both are given equal base current via the 39K s and starting relies on slightly different transistor characteristic or core remanence to get one side  to dominate in the first pulse.
A disadvantage of that is that the remanence of the core may not be consistent every time it is switched off.

Looking at the bad trace, it might be showing that there are 3 short pulses with both transistors on together just before the Idc rises.
And when the Idc rises, there are 6 long pulses with both transistors out of saturation
That might indicate the the core was initially part saturated and the feedback winding was not giving solid +/- pulses.

I think your good trace shows the green transistor conducting first.
To help green come on a bit harder, a suggestion would be to change  39K of the yellow channel to say 56 k and "try another 999 "...