> A partition can’t be deleted until every single piece of data it stores is no longer needed. This is appropriate for some systems, like message queues, where data is needed for a predictable amount of time. However, it’s not appropriate for a workflow system that supports long-running workflows.
If most workflows complete within a few days, we could explicitly copy the long running items forward and then reclaim the entire partition. Sacrificing an extremely good technical solution because we can't be bothered to normalize <1% of use cases seems quite hasty, especially when we consider where this rejection of technological simplicity often takes us.
If most workflows complete within a few days, we could explicitly copy the long running items forward and then reclaim the entire partition. Sacrificing an extremely good technical solution because we can't be bothered to normalize <1% of use cases seems quite hasty, especially when we consider where this rejection of technological simplicity often takes us.