A few reasons:
First it's just boring. 99% of the time, it's mostly boring and what you'd expect. Some dumb Django or Rails app? Who cares. A Python library to talk to some internal service? A backend implementation of some well-known, public feature? Probably knowing about the feature is more interesting if its somehow secret or salacious...
Second the coding environment is very specific. FAANGs have very unique software ecosystems, coding styles, ways of structuring services. So even if you had the code, what could you do with it?
Third, and most importantly, the secret sauce IMO isn't in the code. It's the institutional expertise behind the code.
If you had Google's way of matching ads to search queries, then OK, you have a snapshot of a point in time. Same with Twitter's recommendations. It is moderately interesting and you can learn from it. But these algorithms are _constantly evolving_ around experiments and shipping. The internal teams are constantly learning and evolving how they understand that feature. So the team's mental models about that area of their product is the actual secret sauce.
Frankly if its JUST in the code, like some genius hacker put it there, and nobody else understands or can maintain it, it's not that useful to the organization. Those kinds of bits of code gradually die without a team to scaffold and evolve that code.