Your example is misleading because of your choice of characters (and to some degree typeface), not your choice of alphabet. You would have the same problem (and people have had the problem) using the characters 1, I, and l in identifiers.

This argument does not apply to the original example and it's an odd argument to make.

The biggest downside to using 'φ' over 'phi' is how hard the first is to type on most keyboards, and how it might be harder to google (although I just tested and copy+paste+search gives me the right answer immediately).

I'm a fan of reducing jargon, but this is a very standard symbol and I like its use in a codebase - especially one targeted at people studying mathematics.