Here's an analysis of the Concertina phase splitter using Shekel's method.

The 3 Concertina output impedances (2 driving point and 1 differential) are
plotted versus frequency for equal 100 pF loads on plate and cathode.

The 2 driving point impedances are seen to begin rolling off at around 10
kHz, but the differential output impedance doesn't roll off until about 1
MHz. This is the good performance which the Concertina can deliver if the
loads remain balanced.

Repeating for emphasis--the 2 driving point impedances do in fact begin to
roll off at an undesired low frequency, but not the differential (balanced)
output impedance.

Also shown is the frequency response to plate and cathode with a slight
unbalance in the capacitive loads; 100 pF on the plate and 110 pF on the
cathode. This causes about a 10% increase in gain at the plate relative to
the cathode gain for only 10 pF extra capacitance on the cathode!

And this unbalance in the gains begins at 10 kHz, not 1 MHz! Any
perturbations in impedance or gain due to unbalance will begin at a much
lower frequency than the balanced roll off frequency. Patrick Turner's
advice to add some capacitance to a Concertina would be a good idea for
optimum performance. Perhaps a trimmer cap adjusted for best similarity of
frequency response at plate and cathode.

The tube parameters are more like a pentode than a triode; they were chosen
to emphasize the effects.

The rather tall .gif file can be viewed with Firefox.