I advise you to change the destination location by making appropriate cpan configuration changes, then re-install the modules.   In this way, you allow the installers (and self-test scripts) to “do what they ordinarily do for you, in the usual way.”

This bypasses many potential problems, while incurring only a modest (one-time) cost in time.   Since the code will be installed and self-tested by this process, and because it will install only if all the self-tests pass, it becomes a foundation you can reasonably stand on without worrying (too much) about falling into the river.

(I’m not concerned with how long it takes to build the bridge.   The thing that I earnestly want to avoid is the river.)

When I am, for example, “deploying a server or a web site,” I actually build an installation-script, set up a little “sandbox” somewhere, launch the script and sit back to watch it go.   I want that script to run clean, start to finish, leaving behind a fully configured module library.   If it doesn’t, I fix the script, tear down the entire sandbox (as in: rm -rf sandbox_dir), and try again (and again...) until it works.   So now, the deployed server configuration actually has a completely-repeatable installer.   There are, of course, tools to assist with that process.