We have a large collection on which we have a large Atlas Search index on a Workload Isolation node. In the lower environments we pause overnight to save cost. Upon restarting, the large index has to rebuild, it was taking 10hrs and this was wiping out development time.

In posing this question to support the response was:

"When you pause a cluster, Atlas deletes all data from the Search Nodes. Atlas automatically rebuilds the Atlas Search indexes when you resume the Atlas cluster. While Atlas rebuilds the Atlas Search indexes, the $search and $vectorSearch pipeline stages are unavailable. You can review the Considerations for Paused Clusters here."

So it is expected behaviour. My question is why, it makes no sense to drop an index on pause and have to build it again, can we not persist them?

Support ticket reference : https://support.mongodb.com/case/01449917 (note there is also a poiunt about during the index rebuild none - even small Search indexes were useable)