Transient simulations are perfectly noiseless, so after the engine finds the DC operating point, it will perfectly happily sit there forever and ever, even if it is an unstable operating point (as is the case here).
You can also set initial condition of the capacitor (setting the node voltage means the same thing*), or use different DC operating point methods in the simulation configuration, e.g., set to zero.
*There is a certain advantage to setting the capacitor, even if it may end up being a semantic difference: the capacitor is the only explicit component which stores "state" of the system. That is, the operating frequency is ideally defined only by the capacitor (there are no other time-variable elements), and if you solve the differential equation of a system like this, you must input one other factor, the voltage at t=0, to find a unique solution. Only this voltage matters; all others can be derived from it.
The other advantage may be practical; component "models" that include e.g. ESR, ESL, etc. won't set the SPICE capacitor voltage by .NODESET -- because there are nodes inside that you can't see. You then have to enter the initial condition in the model dialog.
Tim