to shield or not to shiels, that is the question: Whether 'tis nobler in the mind to suffer from not shielding.

In an ideal case there is no need whatsoever to shield a balanced twisted pair. But which case is ideal?

now what makes the case less than ideal
1) common mode limitations: there is no way of predicting how a 485 transciever will behave if a noise spike exceed the common mode limit.
2) slight imbalances: we all (I guess) violate the balanced nature of the pair at connectors, PCB traces etc.. If you have, say, a 10% imbalnace, you will get in trouble if a noise spike is 10 times the noise margin of single ended.
3) component tolerances: there is a slight 'error' in the balanced nature of the transciever, not enought to matter on it's own, but it can break or bear if other imbalances exist.
4) the termination: two points, even if the termination impedance is close (it is never 'perfect') you get some ringing, not enought to matter on it's own, but it can break or bear if other imbalances exist. Also, the 3 resistor termination required if the transciever is not 'state safe' OR the internal transciever circuit to make it 'state safe' will again create a slight imbalance, not enought to matter on it's own, but it can break or bear if other imbalances exist.

Were I to communicate between two distant points in a non-industrial environment and an unshielded twisted pair was already in existence, I would use it rather than insisting on a new cable being pulled.
The reason I use shielded twisted wire is #1 above, even with isolated transcievers there is a common mode limit.

Erik