It's not intuitive, but yes, the grid is a virtual (ac) ground.I don't think the grid fits the definition of a virtual ground... In fact we want it to be quite the opposite, no?
The only Vac there is the error component.
The way that helps me to visualise it is that when considering the resistor chain Rg + Rf (let's have them equal at the vox 470k), at the input of Rg there's say 1Vac and at the output end of Rf there's also 1Vac but of opposite signal polarity.
Therefore the mid point node of the resistors must be null.
Pete
