Audio transformer: What to do with an unused winding?

Option 3 is best

It does not matter whether you connect the primaries together or not, you are still stuck with the primary voltage appearing on pins 3 and 4 as a result of transformer action, and still stuck with that interacting with whatever parasitic capacitance exists.

At higher frequencies, the leakage inductance will tend to decouple winding 34 from 12, so the parasitic capacitance on 34 would be better controlled by connecting it directly to 12. This favours option 3 over option 2 somewhat.

Option 1 does not do what you want. The 150ohm resistor will appear in parallel with the transformed 600ohm load, and present a 75ohm load to your primary.

Generally, for audio, matching isn't really that useful so I'd feed the signal into 1&2 and take an output from 5&6. This means that the 600 ohm load is seen as a 600 ohm load on the 150 ohm source i.e. it lightly loads it. If there is a compelling reason to achieve matching then spell it out.

Which option is preferred? Why?

I'd strongly consider a 4th option. I'd be keen to know that the output is representative of the input so I'd be tempted to use winding 7&8 as feedback to the driver thus, high frequency losses due to leakage inductances can be somewhat mitigated. But, it may not be that important to you?