In my few products' worth of experience doing MOSFET drivers,
I always would minimize (not to say eliminate) shoot-through
in the commoned-output configuration. Eliminating all of it
would add too much to prop delay, allowing too much will ruin
dynamic Icc and overall power-train efficiency. A modest
shoot-through (providing that features are sized for the
time averaged current density, etc.) can be tolerated as far
as product reliability goes.
The point of split outputs is to tailor totem-poled external
FET gate waveforms to prevent shoot-through -there-. Which
is likely more variable and more painful.