It's probably the temperature of glass itself. Speed of sound in solids depends on the elastic moduli (the choice of the modulus depends on the polarization of sound - in this case, it's mostly transverse motion, as you are observing glass oscillation). The change of the density of the hot air may also contribute.
Possible contributions and why they can be excluded:
- density and thickness of glass: thermal expansion coefficients aren't that big (a few $10^{-6}K^{-1}$).
- change of water properties: not really, the part of the glass that interfaces water doesn't vibrate all that much, as water has a lot of intertia.
- elastic moduli: maybe... glasses are known to get softer gradually with increasing temperature. Can't find numbers now, but probably contributes more than thermal expansion.
- hotter air is less dense. So it's essentially the same effect as breathing helium. However, resonant frequency of the glass is not that much affected by the air density (vocal cords are a different thing).
From all this, #3 seems the most likely candidate.
EDIT: now that I think of it, it's possible that while the expansion itself doesn't do that much, stress that is created during nonuniform expansion changes the elastic response (it's pre-stressed, the same reason the glass may break when heated nonuniformly). So... it could still be elasticity, but maybe through a slightly different process than just getting softer... you can test this by waiting for the temperature to equilibrate and trying again.
I'm assuming that you're talking about frequency while keeping the amount of water constant (and the surface at rest).