NVIS really works best with horizontal polarization.
I suspect in the case you cited that the "off-center" fed
T antenna acted like a Windom (which was originally fed
with a single wire) the top of the T was actually radiating
in dipole mode.
While the broadcast band does have the advantage that it
is below the Critical Frequency almost always, the other
problem for propagation is path losses, especially in
the "D" layer of the ionosphere. So the frequency window
where NVIS can be used is limited at the top end by the
Critical Frequency, and at the bottom end by the signal
absorbtion. (This is why 160m, and often 80m, is quiet
during the day, but has lots of signals at night.) Both
of these factors vary with time of day, season, sunspot
activity, etc., so the "optimum" frequency is constantly
changing.
For an antenna, if the existing tower can't be salvaged
somehow (perhaps by rescuing one or two sections and
adding a wire top hat to make up for the missing length)
the I'd consider a dipole or full wave loop. If the
station had room for the radials for its tower, it will
have enough room for either.
There are, though, a number of practical issues that
will probably be the determining factors. First, the
optimum frequency for a station is that stations's
assigned frequency. They probably don't have any other
equipment available. And most AM BC stations don't
have a convenient PL-259 connector that you can hook
your antenna to: they probably use a large, rigid coax
to the antenna (especially if they are running over 5kW.)
Tuning and matching the antenna will be another issue -
you don't just hook and old Johnson Matchbox to the
output of a commercial transmitter. The station engineer
might look out in the Tuning House and see how the
circuit might be reconfigured to suit a different
antenna, but it usually isn't like readjusting the
ordinary antenna tuner. Might be easier to replace the
tower with a kite-supported long wire.
But, IF there was a low-power backup transmitter
available, it could be pressed into service just like
any station for 160m. The ordinary antenna dimensions
still apply. The problem is likely to be finding
enough antenna supports that survived the initial
disaster - I suppose you could disassemble the original
tower sections and them individually for antenna supports
if you have the equipment to put them up.