That is not limited to LEGO: Some researchers from Lancaster University also were researching LEGO bricks (read: Playing around trying to set record-low temperatures for LEGO bricks); and found out they're excellent insulators at sub-kelvin temperatures: https://news.ycombinator.com/item?id=21884755
You can put them in a vacuum. They don't degas much. They're magnet safe and can survive being in really high EM fields. They've got a high melting point. They don't become too brittle at sub-K temperatures. Blocks continue to mate as temperature varies -- Lego's process is fantastic at minimising temperature variation. They're really quite radiation resistant (many plastics fail horribly if you shoot them with >1 kGy). A child (or student!) can make a symmetric structure with them given a vague sketch design. A single "1x" wide Technik piece has a persistence length of about 80 cm. Structures you can build with lego are therefore really remarkably strong.
Every lego is made to insanely tight tolerances to published, well-known dimensions, and, best of all, if they break (an achievement!) you can buy another one for absolutely nothing in science-money terms. (The finance dept. look at you a bit weird when you put fifteen copies of "Spaceman playset" on an expenses claim form though).
I frankly think that most experimentalists somewhere in their life have some lego.
(I have a "Lego MRI" set that is only available to professionals!)
The accelerator I most recently used has a large set of house-grade UPVC glass windows around the target stage to help thermally control it (it's variable temperature). It's after the beam dump and behind quite a lot of lead, but they've been there for a while.
The glass is a weird shade of purple in a pretty gaussian pattern and the formerly bright-white UPVC is brown in places, purple in other places, and looks a bit like wood does when it's rotting and peeling off. If you hit it with a hammer I'm sure it would disintegrate. There are rad hard plastics but again it's all a function of "what radiation", "what energy", "how much" and "how long" and you end up doing monte carlo simulations pretty quickly.