I completely agree with eslaats it feels incredibly hacky that Databricks doesn't support business-hour windows for Continuous triggers out of the box!

We ran into this exact same billing nightmare. Leaving a cluster running continuously overnight just to process a trickle of streaming data was destroying our budget. We ended up adopting the second approach mentioned above: using a lightweight external automation (like an AWS EventBridge cron triggering a Lambda) to hit the Databricks API. It toggles the job's schedule state to ACTIVE in the morning and PAUSED in the evening.

The only tricky part was aligning those API triggers with our actual shifting business schedules. We had to strictly map out our team's core hours to ensure the continuous job was active exactly when the data analysts were actively querying the downstream tables, without leaving it hanging open and burning compute during holidays or partial shifts.

It takes a bit of initial setup outside of the Databricks UI, but using the REST API to pause the schedule state is definitely the most robust and cost-effective workaround until they natively support time-boxing!