That is, the transistors in the array are not connected to each other to implement a specific function.
There are three main motivations for combining several transistors on one chip and in one package:[1] The matching parameters and thermal drift are crucial for various analogue circuits such as differential amplifiers, current mirrors, and log amplifiers.
Often the transistors here are Darlington pairs with a common emitter and flyback diodes, e.g. ULN2003A.
While this stretches the above definition of a transistor array somewhat, the term is still commonly applied.
For instance, for an array of NPN transistors, the substrate must be connected to the most negative voltage in the circuit.