From my view, repos should be mandatory for whatever project and platform, otherwise a lot of issues will arise sooner or later and, automation will never be possible.

Having said this, apart from Databricks official documentation you can get, my experience with GIT repos in databricks is very satisfactory when combining it with other external tools such as Visual Studio Code and Azure DevOps. I'll try to summarize:

Let's say I try to use the best of each tool. Visual Studio Code let me develop code very fast with high quality code control (Pylint, Pylance, etc.), AI assisted (GitHub Copilot, AI agents,etc.), high variety of plugins, etc. while I have the option of executing that code in Databricks directly via extension or by pushing into Azure DevOps GIT repo firstly and then, use Databricks Git repos to pull and run directly in Databricks. All in all, depending on the task I switch from Visual Studio Code and Databricks UI or viveversa but always with GIT repo as common/shared place for code. Besides, Visual Studio Code let me deploy easily Databricks Asset Bundles (DAB) for DEV or TEST environment by running proper commands in integrated Databricks CLI tool. This is only my personal opinion, I like to develop most part of code in VS Code and then, switch to Databricks UI to test it in personal/shared cluster, instead of running it directly from VS Code (it's much slower in my case and could spark security concerns in corporate scenarios).

In my case, Azure DevOps is used to CI/CD purposes. It features very straightforward tools to integrate Databricks CLI and deploy DAB in a very simple and secure manner. Very important as well, Azure DevOps is the tool to run and review Pull Requests as Databricks repos is not offering that yet.