Underrated transistor?

[earckens]

In the schematic attached there is transistor Q5 (C1815 SOT-23, spec sheet attached) that in theory should withstand the conditions to which it is subject:
Vcb max 60V, Vce max 50V, Ic max 150mA, Pc max 200mW.

When running this pcb in test modus, with a small 1000µF 63V condensor, all runs ok, hours on end.

Yet when even one single 4700µF 63V condensor is connected, the circuit output goes into oscillation after some time (ie output voltage fluctuates +/- 5 to 10V), and, as was the case just now, suddenly burns out Q1 (and takes the LM723 with it).

Condensor Vcc = 46V.

What could be going on, except for the fact that in theory things should be fine?

How to remedy?

EDIT: C7 has been replaced with a 100µF condensor, the 18Vdc line is free of any ripple.

 

[alec_t]

What could be going on

Just a guess, but if the power supply switches off (or goes into violent oscillation), Q1 base voltage may drop to 0V, in which case Q1 Vbe will be -7.5V. The maximum reverse voltage of the base-emitter junction, per the datasheet, is only 5V, so perhaps the junction breaks down.
Given the values of R12 and R15 the breakdown current would be small, so whether or not it would be fatal is debatable.

 

[earckens]

Just a guess, but if the power supply switches off (or goes into violent oscillation), Q1 base voltage may drop to 0V, in which case Q1 Vbe will be -7.5V. The maximum reverse voltage of the base-emitter junction, per the datasheet, is only 5V, so perhaps the junction breaks down.
Given the values of R12 and R15 the breakdown current would be small, so whether or not it would be fatal is debatable.

That sounds plausible, but where does the -7.5V come from? The zener diode is connected to Q3, while the burnt out transistor is Q1?

And I cannot find a transistor with Vbe better than 5V?

Even a BD139 (spec attached) does have 5V Vbe only?