In an audio amp, for instance, you basically have a transistor connected to a power supply rail and a very low ohm fusible resistor, e.g. 0.22 ohms.
Biasing sets some fixed current to flow through this resistor with no signal.
Now, suppose there wasn't any way to limit the the Vbe drop. Once the transistor starts to heat, it conducts more, thus increasing the current through the fusible resistor. Now, there are more electrons available because of thermal energy, it heats up even more. So, there is your thermal run-away problem.
---
A lot of audio amps has what's called a Vbe multipler. This multiplication is adjustable. If it is too high, the current through the fusible resistor will continue to increase until something breaks.
I built this amp (The Leach Amp)
https://users.ece.gatech.edu/mleach/lowtim/ in the mid 80's with my own design changes. Here
https://users.ece.gatech.edu/mleach/lowtim/2ndstage.html is a discussion of the Vbe multiplier.
My changes were mostly power supply related and component selection and protection. Nearly all the resistors are metal film. The bias adjust is 10 turn. It's packaged in a 2 RU case.
The power supply is 4 x 35 VAC at 3a each with 9600 uf of computer grade caps on each rail. The transformer is a custom Torroidal.
With ~40,000 uF at 50 V capacitance at turn on, there is quite an inrush current. An inrush current limiter was built into the system. The design is such that reversing the NPN and PNP output transistors will result in one fried resistor. With a fancier design, I could even stop that from happening.
There is no inherent VI limiting or other protection mechanisms except a 3A AGX fuse at the output and a four 3A 3AG fuse, one on each power rail. If one rail goes down, all rails go down and that magic resistor pops.
The AMP doesn't have a power switch. It's turned on by another device that uses a triac.
The audio media these days sucks and can't do the system justice.