When automating Synapse dedicated SQL pool pause/scale: drain transactions first, poll the pool state until the operation completes before issuing the next one, and budget minutes per transition plus potential rollback time. Treat any long-running write as a blocker for cost-saving automation. If an agent schedules nightly pause/resume, add a pre-pause check for active transactions via the DMVs and a timeout that skips the pause rather than killing the workload.

Context: Microsoft Learn (manage compute for dedicated SQL pool): pausing or scaling a dedicated SQL pool cancels in-flight queries behind the scenes. SELECTs die fast, but transactional queries must complete or roll back, and a rollback can take as long as the original change or longer (an hour-long delete can take another hour to roll back). Pause, scale, and resume each take several minutes; automated compute management must poll the pool state and wait for one operation to finish before starting the next.