It depends essentially on a few considerations:
- What's the available power supply voltage difference across the circuit?
- What's the input signal voltage range?
- What's the gain required? (Or, what's the output signal voltage range?)
- What level of signal distortion is permitted?
- What's the circuit's operating temperature range?
- What's the design range for transistor parameters used in the circuit?
Obviously, if the output is DC coupled to the next stage then the rule is tossed out on its head because the next stage's requirements determine everything and the rule of thumb becomes mostly useless. Let's discount that and assume the output is AC coupled.
Things are less critical when the available power supply voltage difference across the circuit is very large. In such cases, the simple rule can be applied without thinking much. This is because there's lots of working headroom for both the collector and emitter. The emitter headroom allows significant compensation for operating temperature variations, so that's no longer much of a problem. The collector headroom helps to minimize collector current variation and thereby reduces output distortion and assuming it is also large with respect to the output swing these two issues, distortion and room for the output, are no longer much of a concern.
So large power supply voltages sweep away a number of concerns and the rule can be readily applied without much concern. That doesn't mean it has to be used. But it can be without getting bogged down with other problems listed above. And it has the advantage of being easy to apply without risking the remaining considerations.
It's in the case where the available power supply voltage difference across the circuit is small that the rule should be thrown out -- or, at least, set aside while other concerns are examined more closely.