Audio Designs have historically considered the dampening factor or ratio of load impedance to source impedance over the signal bandwidth. At low frequencies the source impedance is limited by the sum of all source ESR + storage capacitance divided by the feedback gain. At low f, the load impedance tends towards the sum of all ESR's ( speaker, wiring). In this case , the ripple current passes thru both source and load and if the capacitance of the supply is sufficiently large, the voltage ripple becomes the ratio of source to load impedance. Feedback is used to lower the source impedance and where feedback is taken affects this dampening factor (DF) as well as phase margin, (PM) if reactive affects the phase of feedback. If one considers the ESR of the bass speaker coil is a small fraction of 8 Ohms and a typical dampening DF ranges for quality designs ranges from 100 to 1000, the cost vs power losses is a significant tradeoff.
Thus the cost of a full vs half bridge must weigh these factors which ends up being THD and Ploss vs cost.
The cost vs complexity vs THD and P.out levels will determine if it is a trival cost. Essentially there are 2 stages SMPS regulation. THe source supply voltage and the signal voltage. For the same P.out, a full bridge (FB) can reduce the supply voltage required in half while doubling the signal voltage, while dioubling the supply current and expected losses and halfing the signal driver current and losses for the same MOSFETs. Generally a full bridge also doubles the switch losses, but a lower Vds rating tends to also lower RdsOn at the same cost to a lesser extent so overall it is not quite 2x. So it becomes a tradeoff between power loss, dampening factor, cost and output power, which in high volume may not be a trivial optimization. The back EMF from the speaker coil and series choke both must be considered for source ripple or so called pumping from the dynamic signal load at low f.
Considering a buck regulator efficiency tends to increase with higher Pout design criteria and 90~95% with a good design is far greater than any linear design at 50~65%, Class D is pretty good, but not perfect, so other classes were created ( eg Class E,F,G...) to improve performance in these key areas ( $, THD, efficiency.) but gain margin is also a consideration with choices for feedback ( voltage, current, before/after filter).
The Full bridge has a this advantage of back EMF cancellation but you also have the choice of reducing source impedance optimizing the duty factor, d.f. of the deadband time and commutation period, which raises the source impedance slightly by (1/1-d.f) This required deadband commutation for an active bridge is open to prevent shootthru currents for some X microseconds and a full bridge will have 2 deadtimes per cycle or twice the potential ripple that must be attenuated at high f.
At any frequency, lowering ripple, Ploss and THD, ultimately becomes a goal to raise the dampening factor over the measured bandwidth.