How to combine both MDAPI and SFDX way of working?

Currently we are looking into adopting SFDX on our enterprise org.

A brief overview of our setup/current status:

• "Monolith" org
• Everything is under version control (GIT) - MDAPI format
• Working pipelines for all higher test environments (full copies) & production

Since we are way above the limit of 10K components & have an existing operational projects running in parallel, we cannot start with migrating everything in a big bang to SFDX.

In order to do this, we were thinking of starting small, with a specific team on SFDX and all other teams still utilizing regular sandboxes. The following approach seemed feasible to us:

1. Convert the MDAPI repository to Source and create a project specific branch out of it
2. Create a feature branch for the developer to work on (push specific code to scratch org and pull back if needed)
3. Once finished, merge feature branch into project branch (still all Source format)
4. Convert Source to MDAPI and merge with full org MDAPI code base

The structure of the branches would be the following (see attachment)

This now brings me to the core question, what is the best way to handle step 4? I was thinking to do the following:

--> Create regular sandbox, merge code to this sandbox and retrieve full MDAPI code?

What do you guys think & most important, how can this be automated?