It takes approx. 0.2 seconds for the oscillation to build up to a significant amplitude. The loop gain is initially greater than unity so the oscillation grows exponentially until Q2 starts saturating in the troughs, when the resulting clipping reduces the large signal gain to unity.
If you add the startup option to the .tran command, (check 'Start external DC supply voltages at 0V'), it adds a power-on transient which kick-starts the oscillation so it stabilises in under 20ms, however the downside is that the .op operating point is now useless.
A technique that may be useful with very high Q oscillators to avoid unreasonably long sim runtimes before the oscillator noticeably starts, is to force them at or very close to their operating frequency with a sinusoidal current source for a number of cycles to build up the amplitude more quickly. Once Ncycles is reached, the current source output becomes zero, effectively removing it from the circuit. The forcing current should typically be an order of magnitude less than the AC component of that due to steady oscillation at the point of injection.
To zoom in on a waveform, drag a box over what you want to see zoomed, then right-click and 'Autorange Y-axis' to best fit the amplitude of the zoomed in trace(s) (or not if you want to Zoom amplitude as well). To get back to the original unzoomed trace, right-click and 'Zoom to Fit'.