Essentially the same John suggested. I chose at first to leave the transistor to auto-oscillate. I was able to trim the supply voltage, so I noticed that there are regions in which the pulse width was rather cahotic, while in other case the pulse repetition rate was more regular. In my circuit, I used a coaxial cable to store the charge and not a capacitor. For the PNP transistor, I just reversed the circuit.
If you bias the C/E just below avalanche, an impulse on the base can trigger it. I made a few tries, having a look at the Linear AN I mentioned above and it seems to work.
You have to try. These transistors are not designed to work in the avalanche region. Some will avalanche nicely, some simply do not. Be extra careful with the fast scope: start with a big attenuator with a very good frequency response and then decrease the attenuation if you need. Sampling head are fragile.