a very interesting post - I have for a long time looked out for any hard information on what actually went wrong with the radios in this case. Usually crystal failure is blamed - but I think we are getting nearer the truth here.
You mention the WS22 and WS68P - these are both sets I have heard before as being used and not working properly. In addition to the problem of limited ground wave range - which I totally agree with - I can offer another couple of thoughts:
1. The 68P's frequency coverage stops at 2.9MHz - the 22 lower end is 2.0MHz - this gives them only 900kHz in common. They may have struggled for an empty channel. Interference can reduce range very severely.
2. The WS22 looks like a WS19 to the uninitiated - but it has a difference in the tuning procedure which can be fatal to the untrained. The WS19 generates its transmit frequency from the Rx local oscillator, and has one tuning knob for both Tx and Rx, thus making it a true transceiver. The WS22 looks the same - but has a free-running MO for the Tx, which is only roughly tuned to the Rx frequency by the single tuning control. An extra "Tx trim" knob was introduced to net the Tx to the Rx. If this 2nd netting step is not done - then the Tx can easily be 100kHz+ from the Rx frequency - and all comms will fail.
I have it on hearsay evidence only that the radio ops at Arnhem had not been specially trained on the WS22, and fell foul of this quirk of this set.
I regard the WS22 as exceptionally poorly designed - introducing a set that looks very like another existing set (the WS19), but having a slight but critical difference, which can fatally upset operation should have been caught - but wasn't.