Whether it has an effect is not an issue. The issue is that it is referentially transparent.
Sure, we could have a discussion about whether partiality is an effect. Someone, somewhere is having that discussion. If the Haskell committee had had this discussion in 1987, chances are we wouldn't have Haskell at all.
This is the exact same objection people have been raising for 20 years. Yes, lifted types and partiality are probably bad. No, this isn’t an argument against distinguishing between functions and other objects. If anything, it just goes to show that Haskell should have been even more extremist about effect management.