The complication comes from solubility of the solvents with each other whilst simultaneously having some solvation power of a solute. I have made a compound before that was so water soluble that it caused sodium chloride to precipitate out. A real example was of an organic compound that had 1%w/v solubility in ethanol but around 6.5%w/v in toluene. In a 50:50 mixture, the compound had solubility of 13%w/v. A completely non-linear result! The solubility was even higher in 90:10 toluene-ethanol. So, the complex nature of three way interactions mean it would be very difficult to predict these solubility graphs just using two data points of the two pure solvents.
If we asked the same question of boiling points, we would point straightaway to azeotropes. So, analogously, other properties exhibit this behaviour. However, in this instance the nature of the solute as a third component makes it too complicated to predict without having the wealth of property data available for all components. If this is known and you have the software, you may have a chance.