... Q4 and Q8 are both set up as current sources and they are fighting each other trying to drive different currents through the same wire between them. Output voltage will be determined by the difference between their collector currents multiplied by paralleled output impedances of the two transistors.
@magic, is that your insight? If so, my congratulations! I ask because I have been using for a long time this functional approach to explain the idea of
dynamic load... but no one has accepted it so far. For the first time I see someone using such an explanation. Here are some related links:
https://www.circuit-fantasia.com/my_work/conferences/cs_2005/paper1.htmhttps://www.circuit-fantasia.com/my_work/conferences/cs_2005/paper2.htmhttps://www.researchgate.net/post/Can_we_apply_input_current_to_the_collector_of_a_BJT_whose_base_is_held_at_a_constant_voltage_and_to_take_the_collector_voltage_as_an_outputhttps://www.researchgate.net/post/Where_is_the_dynamic_load2Structure. The circuit represents an AC common-emitter stage with dynamic load. For the purposes of understanding, we can think of this arrangement as of two "fighting" current sources (current-stabilizing non-linear resistors)... or a source (Q4) and a sink (Q8). They are "incorrectly" connected in series so each of them tries to pass its current through the common path. As a result, a
current conflict appears between them, and the voltage of the middle node vigorously changes. Similarly, we can think of a BJT differential pair (aka
long-tailed pair) as of two "fighting" voltage sources that are "incorrectly" connected in parallel so each of them tries to set its voltage at the common emitter node. As a result, a
voltage conflict appears between them, and the current vigorously steers between transistors.
Operation. Initially, we have to adjust the currents produced by the two sources to be equal so the output voltage at the common point to be V2/2. We can control the output voltage by changing the current of the one of sources (while keeping the other constant) or both. We can do it by means of the voltage divider R1-R2 (DC) and V1 (AC) for the sink Q8... and by the current-setting resistor R3 (DC) for the source Q4.
We can imagine the circuit operation in terms of static (instant, chordal) collector-emitter resistances instead of currents flowing through them. This means to think of the two collector-emitter junctions (CE4 and CE2 in the attached picture) as of two partial resistances (RCE4 and RCE2) of a potentiometer as shown in the attached picture. In the graphical representation, static resistances are represented by lines starting from the origin of the coordinate system (the picture shows the more sophisticated case when both elements change their resistances in a differential manner).

Initially, the static collector-emitter resistances are equal... i.e., the potentiometer slider in the analogy is in the middle. When the input base-emitter voltages (VBE4 and VBE2) change differentially - e.g., the magnitute of VBE4 increases while of VBE2 decreases, RCE4 decreases but simultaneously RCE2 increases... like the two partial resistances of the potentiometer when moving the slider to right. But the total resistance RCE4 + RCE2 remains constant... so the common current flowing through the network remains constant as well.... only the output voltage VA vigorously changes.
In the graphical representation, RCE4 IV curve rotates clockwise... and RCE2 IV curve simultaneously rotates in the same direction... so the operating point A vigorously moves to right along the horizontal blue line. When VBE4 and VBE2 change differentially but in the opposite direction, the processes are reversed...