I work on large-scale systems software, mostly high-performance database and analytics products. Docker's sole function in that kind of environment is easy replication of dev and functional test environments. No one deploys this type of software in Docker or similar due to the performance implications. Docker was a convenience, it wasn't providing anything critical. I replaced it with a small set of shell scripts and similar.

For this type of development, the number of external dependencies you have is very small. Aside from the compiler, build system, and testing, the number of third-party libraries linked into a release can usually be counted on one hand. Again, for good reason.

Basically, I am doing what I did before Docker, writing a small set of scripts that can perfectly reproduce the environment where I need it. This is made much easier because modern build systems have good dependency management built-in so writing those scripts is a lot less involved.