You're trying to add a massive amount of complexity and new technologies. That makes it much more likely this process fails, and is the opposite of what a proof

You're trying to add a massive amount of complexity and new technologies. That makes it much more likely this process fails, and is the opposite of what a proof-of-concept will give you. A build system like bazel or pants is not really related to deployment. You're overthinking this. As alternative, try to do a small proof-of-concept: take one feature from your monolith and move it out. Can use PHP, doesn't have to be new language. Don't pick tools in advance, instead just go and do it and pick tools based on the problems you encounter. This will give you a much better and realistic sense of what problems are relevant in your particular setup. And will also make it much more likely your project will succeed and get management buy-in.