Heterogeneous task retry requirements cannot be satisfied with global scheduler settings

When a Luigi workflow contains tasks with different failure recovery characteristics (some needing hourly retries, others needing 10-minute intervals, some requiring disable-after-first-failure), global scheduler-level retry configuration forces a uniform retry policy across all tasks, leading to suboptimal retry behavior: either too aggressive for tasks unlikely to recover quickly or too conservative for tasks that could succeed sooner.

For Luigi versions before PR #1791: Split workflows into multiple smaller executions using ExternalTasks, each with appropriate retry configuration for that subset of tasks. For modern Luigi versions: Use task-level retry policy configuration to set per-task retry delays, disable-num-failures, and retry windows.