You are right, sorry. I have overlooked the resistor R2 that - indeed - establishes negative feedback (in a relatively uncommon configuration).
Nevertheless, for the opamp as well as for your cicuit the voltage difference between both inputs never disappears.
That means, the concept of "virtual ground" is only an approximation and a simplification for calculating purposes.
Due to the negative feedback (NOT for pos. feedback) the difference is determined by the open loop gain - resulting in some microvolts in case of opamps.