Hard to tell from the illustration in the OP link what the final "focus
plane" will be, as it just looks like most gimbal mounted telescopes. The
three mirrors just focus the field of view. I took my cue from:

"This means that its field of view will span 4° - equivalent to eight full
moons - compared with the 0.1° seen by other large telescopes." and "image
the entire sky across three nights, producing an expected 30 terabytes of
data per night"

Satellite bourne earth sensors have a field of view (FOV) of about around 12
to 15 degrees (Landsat 3, Landsat 5) from an altitude of about 800km,
compared to airbourne scanners with a FOV of up to 80-90 degrees required to
get a reasonable
ground swarth width from airbourne height, say up to 70,000 feet (U2).   My
reading of the 4 degree FOV of this telescope is that it is not getting a
whole sky view instantaneously, hence the 3 nights for data collection.  30
terabytes of data depends to some degree on the spectral / radiometric
resolution of the sensor. That is how many bits the data is collected at for
each pixel and how many bands it collects for each pixel.  Landsat 5
typically 8 bit 7 bands. (AVHRIS has 224 bands). Full scene image of 185km x
185km is about 5286 x 5286 pixels square (raw) 7400 pixels map corrected
(resampled).  The data quickly adds up in volume. I vagely remember a
reference to Australias daily weather satellite photo (1km pixel resolution)
equals about 3 Terrabytes a year, so the 30 terrabytes over 3 nights may be
in the ball park, it is a big sky  :)

The gap issue is no issue. There are many remote sensing platforms today
that use broom and or multiple sensors. Landsat 7 uses 16 for each pixel.
One is designated a reference and the others adjusted to comply. This
ensures redundancy in a platform which is hard to get at to maintain.  Even
so it does have a duff sensor (died after it was put in orbit) which is
radiometrically corrected by adjacent pixel samples. The missing row is zero
written then an algortihm applied to sample adjacent pixels, average and
fill in the missing data. Digital cameras have dead pixels in the sensors
from manufacture, these are "mapped" out in the camera at its time of
manufacture in a similar way.

I am a little sceptical like you, I find it hard to envision the manufacture
of a 3b flat plane sensor. I just know that technically it is easier to
build a remote sensor in the same vein as typical geospatial scanners
already in use. I suspect the "3b camera" is no more than an earth mapper
turned skyward, with a bit of journalistic licence  ;)

For now I will just have to wait for more detail from the source....