I do did the similar kind of work in my recent project, where I need to run many SQL DDL's , so I automated the process using databricks jobs, capturing the dependency using a metadata table and creating tasks likewise in job through job api's, doing so we reduced the runtime and resource consumption significantly by 33%.

So, Instead of creating separate workflow/job for each priority task which would result in too many jobs and not recommended too, we can create a master workflow/job which captures table according to priority orchestrates and passes them to child workflow where other filtering happens and passes the list in for each task where incremental ingestion happens is the best approach and also make sure to provision the job cluster with required compute memory, so that it'll not take so much time to compute.