The main benefit that I imagine is, ability to break the
classical linkage between small signal stability and high
load-step slew capability. With an all-linear loop there
is a pretty fixed relation. But a loop that is (say) high
gain low bandwidth near null, but lower gain / higher
BW as you slide away from setpoint, can perform much
better in applications where the load current is highly
and abruptly variable. I know older generations of the
Vicor converters did this, "linear style" but with a split
loop - one "DC" amp and one "transient" amp. Of course
a digital control core offers arbitrary transfer function
capability (provided it can keep up). And you'd need some
way of proving (formally, in many critical-application
markets) that the design is indeed stable. That is one
reason why the traditional designs persist, they are so
well understood and documented. There are plenty of
academic schemes which are touted as great (for
something) but will never amount to anything because
they are not OK for everything (list of must-XYZ for
acceptance).