It's some kind of self-oscillating converter, yes. R905 gives positive feedback, making the Q901-Q903 area a power Schmitt trigger (with the current capacity of Q901!).
It's not going to be very fast, because the transistors are only turned off with base-emitter resistors. (Q901 is wired as an emitter follower and doesn't fully saturate, so at least it won't go terrifically slow.)
It won't be bistable, because the inductor will eventually charge up with more current than the circuit can handle, pulling it off. That is, if current rises to a few amperes, the switch node's voltage is pulled down, turning off Q903, and then Q902 and Q901. This incurs a large gulp of switching loss, as Q901 would dissipate on the order of 4A * 160V for a few hundred nanoseconds, and Q902 a smaller amount.
Anyway, bistable operation wouldn't make sense, because there's nothing else in the circuit to switch it on and off.
I don't think it will be monostable either, because when the transistors turn off, the switch node voltage falls slightly below GND (clamped by D991). When inductor current reaches zero, D991 begins to reverse bias. But it doesn't turn off instantly, it remains on for ~100ns (reverse recovery time). This causes the inductor current to reverse, so when the diode does turn off, the switch node voltage flings upwards again. Which turns on Q903, and so on.
So, without plugging it in or simulating, my guess is it's an astable free-running oscillator, with the time constants defined by the inductance and the transistors' operating point, and maybe C907.
Q904 is the error amplifier. R908-R910 are the voltage sense divider (and R912, too), ZD902 (plus Q904 Vbe) is the reference voltage, and Q904 collector is the gain node. C907 may be for compensation, but the loop gain isn't very high anyway (maybe 20?) so it's probably not terribly unstable to begin with.
R902, R906 and R907 seem too large to do much, compared with the small value of R905. (R905 looks drawn in. Is it the correct value?) If as shown, then, R905 will keep Q903 quite close to the switch node voltage. In the 'on' state, D902 is reverse biased, and Q903 (base and emitter) is up around, say, 150V. In the 'discharge' state (D991 conducting), D902 is forward biased and ZD901 is dropping ~12V, keeping Q903 emitter around (B+) - 13V, so it will E-B avalanche with a rather large current (50mA?), which isn't good. Yeah, I don't think that resistor value is quite right.
Also, that should probably be D901?
Anyway, the overall operation is probably much like an astable multivibrator, with a bias tweak to its feedback arm, so the frequency and duty cycle are under control. Sort of a simplified version of modulating a 555 timer with pin 5.
Whether this wall of paragraphs actually helped your understanding, I don't know. In any case, check the most likely suspects: cap ESR, blown transistors, burned or out-of-spec resistors.
Tim