From a DC bias point of view, short out the base resistors and its a simple two transistor current mirror. Q1 would then be diode connected and the control current of slightly under 1mA flows through it. Adding back the base resistors doesn't stop it acting as a current mirror, it simply increases Q1's collector voltage slightly, such that there is Rb*Ic/hFE across the base resistors, but Q2's base resistor prevents Q1 shorting out the AC signal. As you have reasoned, Q2's load resistor is half of Q1's current set resistor, so will have half the supply voltage across it (less about 0.3V), hence the nominal 10V bias point of Q2 collector.
Of course, hFE is a fiction as a BJT is actually a voltage controlled device with its control input shunted by a diode junction, but as datasheets typically don't provide Gummel–Poon model parameters for transistors, unless you are lucky enough to have a manufacturer provided SPICE model, or are prepared to spend the time to curve trace a statistically significant number of devices, then curve fit the Gummel–Poon model to the gathered data to get the model parameters, the best one can do is roughly estimate the operating point then scale off from the datasheet's hFE vs Ic and Vbe vs Ic graphs (assuming you have a comprehensive datasheet . . .) and iteratively refine it.