There’s nothing wrong with writing separate automations - if they’re well named, annotated and labelled maintenance and troubleshooting are much easier.
Loops are to be avoided because they extend the running time of the automation. If an automation is interrupted (by HA restarting, for example) it will not resume until the trigger fires again, and then it will start again from the beginning. The same applies to delays of more than a few minutes.
Have you read this?
It shows that what you’re aiming for is achievable, but the author is an AI professional and I believe he has spent quite a lot of time and money on the project. 