[Best-Practice] Using yarn in production, using cache to populate `node_modules`

Is it a good practice to use yarn in production?

Cons:

• docker best practices suggest to create prod images as small as possible, which would mean excluding yarn from the image - It's also an additional possible point of failure in your app
• In case of monorepos you're adding the deps of all the other repos (unless you carefully use pnpIgnorePatterns)

Pros:

• the fact that dependencies are zipped translate to a smaller bundle size.

In case I want to stop using yarn in production, is there a way to populate node_modules from the .yarn cache, without passing through the registry?

Start by reviewing Yarn's production-install behavior, the .yarn cache, node_modules population, and Docker image practices described in the issue. Document whether production can avoid Yarn and the registry while using the cache, including any monorepo constraints; no files or tests are named.