Databricks Jobs `for_each_task` accepts exactly one nested task. For workloads that run the *samemulti-step pipeline over many datasets* — a very common metadata-driven pattern — this forces achoice between two architectures, each of which gives up s...
