It’s 189 sensors from two different manufacturers built in 9-sensor rafts, which are tested before integration into the full focal plane, which is then tested again.

Then you can do pinhole images before eventually integrating in with the larger assembly.

Doing so requires more than just specially designed lenses and sensors, Utsumi says, because no matter how well designed and built those components are, there will be imperfections. For instance, consider an image taken by an ordinary camera: There will always be some distortions in shape and color near the edges. There will also be slight distortions in the digital sensors as well, and similar effects will hold true for the LSST Camera. "We need to understand what's going on there so we can correct for it."

Utsumi and his team took thousands of images over three months with the LSST Camera sensors of all sorts of shapes and patterns. They then compared the camera's images with the originals to understand how to correct for any distortions or errors. The team has also worked on how to correct other issues, such as the fact that brighter objects appear larger than they actually are, as well as "ghosts," or images of an object that appear because of electronic crosstalk between sensors within the camera.

from https://phys.org/news/2023-09-lsst-camera.html

For the characterization of CCD image sensors, it is the same idea, see for example this company that does CCD testing machine: https://www.ci-systems.com/ccd-testing

For the mirrors, I once visited a lab that does mirrors for astrophysics, and it measure the mirror shape with wave-front instruments (basically it's interferometry, you look at the phase difference of a wave to measure distances).