Hi,
The only way that I think you can do this is to seperate out the Materialized views from the pipelines into standalone SQL jobs with a TRIGGER ON UPDATE clause. This means they only run if the streaming tables are updated. However, this approach is fine if you just have a few MVs but if there are lots and there are many interdependencies it won't be good to manage. I think the thing to consider is that MVs only process incremental records so if the records in the source tables are changed the MV process kicks in but it will barely use any compute as it won't be doing anything.
I hope this helps.
Thanks,
Emma