Managing Odoo through helm works absolutely fine... until you introduce a high throughput system, or the job queue.
At this point standard helm functionality is started to be stretched: you need to be able to do things like scale down deployments, wait for upgrades to complete, etc.
You can implement init containers, upgrade hooks, etc. but it seems like an operator becomes a better fit the more complex the needs of the installation.
This is a WIP operator that I am working on in my own personal time as an experiment.