detector on detector seems to spuriously set B1 or B2 or at least
 panic DR when we honour it

point setting doesn't seem to work
 residual stuff from CDU is present and changes pt0 when you
  * turn on
  * say `+' to master to turn on CDU
  * switch to programming mode
 pt0 on pic2 is the one nearest the switch box, moving to hacker's right
 (we think - I haven't checked this yet)

crash readout: master i2c should go into known state

layout polarity diagram colourful segment encoding:
 make moveable feature selection in subseg2display bitwise rather than
  setwise

safety system no crashes at crossovers please