Then how about running a low resolution like the 30 lines of Baird's mechanical televisor? That could scan slowly enough to have the beam be seen, remember that 30 non-interlaced was chosen as a compromise between detail and flicker on the rather slow mechanical system/
Think about it this way: The faint light of the electron beam through the gas is already faint as a single beam.
Now you are going to spread the same amount of light energy over a larger area as you pass the deflection coils. The same amount of photons divided by a larger area, makes the brightness incredibly low.
If you increase the beam current, you'll be able to see it. That might actually be necessary to make the image on screen bright enough, because the dot of light will also be spread over a larger area. If you adjust the dot of light on an XY scope with no signal to normal brightness and then apply signal, the trace will be very faint.