A bridge-tied load (BTL) amplifier circuit doubles the output voltage (and potentially quadruples the power, output current limits notwithstanding) by using one "master" opamp for gain followed by (i.e cascading its output into) another "slave" opamp with unity gain but used in inverting config. This is type of cascading BTL circuit is illustrated for example in Fig. 1 in Maxim app note 1122 and in Fig. 73 (on p. 24) in the OPA454 datasheet (OPA454 being TI's highest voltage opamp it is somewhat natural that its datasheet is replete with circuits for increasing output voltage swing even further.) Not all opamps are unity-gain stable (unless externally compensated), but it is possible to adapt the cascading BTL circuit for opamps that aren't unity-gain stable (without adding external compensation) by feeding only a portion of the master opamp's output voltage to the slave opamp through a simple voltage divider. For example, Marston's Audio IC Circuits Manual shows such a circuit using TDA2030 (which is not-unity gain stable) in Fig 4.28 on p. 108. So lack of unity-gain stability of the opamp is not a show stopper for the cascaded BTL approach, although it does increase component count slightly.
On the other hand, a non-cascading BTL circuit can also be built simply by feeding the input signal into one inverting and one non-inverting opamp in "parallel" [sub]circuits and adequately changing the gain (sharing it) between opamps. According to the aforementioned Maxim app note 1122, which shows such a circuit in its Fig 2., that's actually a better approach than the cascaded BTL because "Both [opamps] amplify the same input signal, so neither one reproduces noise, distortion, or clipping introduced by the other". According to the same Maxim app note, a non-cascaded BTL circuit has additional advantages in the case where the availability of a single-rail power supply (such as car audio etc.) dictates that the opamp inputs additionally need to be DC biased (for AC-coupled [that is capacitor-coupled] inputs); chiefly among the advantages claimed in this single-rail supply case, it is said that the non-cascaded BTL circuit "eliminates signal injection into the high-impedance bias source".
After having heard all praises lavished on the the non-cascaded BTL topology in Maxim's app note (that other sources seem have missed), my question is: what are the disadvantages of the non-cascading BTL circuit relative to the more typical (judging by the sources I've surveyed) cascaded BTL? (And, sub-textually, why is the typical BTL opamp-based circuit exemplified using the cascaded approach?)