Should CrossHair diffbehavior detect reordered parameters?

[pschanely]

The TIM Improves Merging project notes that crosshair diffbehavior always supplies the same arguments in the same positions, even if e.g. the argument names have been swapped. (see additional discussion in the report PDF).

It's not fully clear now far to go here, and what balance to strike between being smart and magical/confusing. What if the arguments are slightly re-named? What if the type is changed? Let's collect ideas here.