In an ideal Black-Scholes setting, the Risk-Neutral measure $Q$ is unique and so, obviously, does not depend on what derivative instrument we want to price.
Assume some deviation from perfect markets (e.g. trading costs), and also assume that the risk-neutral pricing formula is still valid: $$\text{Price}=E^Q[\text{future payoff}].$$
In this case, it seems that $Q$ may be option-specific, even though the underlying asset is the same. In particular, if replicating one option has different expected trading costs than another option, then it could be that to price the two options, we need to use two different measures $Q_1$ and $Q_2$ in the above formula.
Is there any formal analysis to support (or not) the above reasoning?