They weren't 'added' they were left out of the first poor circuit, which is probably why it died.
The smallest ones balance the parallel transistors, and are absolutely crucial - the others are a good idea, though not as crucial - but they do help to speed the switching.
My circuit worked very well for years. It was deleted and replaced by the "correction" made by moderator (MP) at Electronics-Lab:
1) His circuit does not have enough output transistors for 500W output. They will smoke and burn.
2) His output transistors do not have emitter resistors so the transistors must be matched.
3) It does not have the protection diodes. When a positive voltage spike on half the center-tapped transformer winding causes a negative voltage on the other half of the winding the diode conducts in series with the forward-biased collector-base of the output transistors and clamps the voltage spikes.
4) The resistor in the CD4047 oscillator has a value that is way too low which required the capacitor value to be way too high. The frequency is then wrong and is different for each cD4047 IC. The datasheet for the CD4047 recommends a minimum timing resistor of 10k.
5) Two quad opamps were used instead of one dual opamp.
Ronnie has used my version for years and improved its reliability by adding a zener diode for the CD4047 IC. He found that the output voltage was low when the load is 500W so he used a 10V-0V-10V transformer to boost the voltage a little.
3) It does not have the protection diodes. When a positive voltage spike on half the center-tapped transformer winding causes a negative voltage on the other half of the winding the diode conducts in series with the forward-biased collector-base of the output transistors and clamps the voltage spikes.
The diodes stop the base-emitter of the output transistors from having avalanche breakdown and limit the positive voltage spike at the collectors on the other side.