Downstream teams should have already been notified about this change on PM-2742, but we are erring on the side of over-communicating the change.
As of 6.0.0-rc0, the server will permit spilling to disk during query execution by default. If users want to explicitly prohibit disk use, they can pass allowDiskUse:false alongside find or aggregate commands.
This project also introduces a new setParameter called allowDiskUseByDefault. This parameter applies only to mongod, not to mongos. As in previous versions, query execution on mongos will never spill to disk. The default value of allowDiskUseByDefault is true. Operators can switch this parameter to false at startup or runtime on a per-node basis in order to restore the pre-6.0 behavior of prohibiting spilling to disk by default.
In addition to find and aggregate operations, mapReduce commands, count commands against a view, and distinct commands against a view will respect the value of the allowDiskUseByDefault parameter.
Currently, mongodb queries run with 'allowDiskUse' set to false by default. This causes queries to fail when a blocking operator exceeds a certain memory limit, rather than spilling to disk.
It would make more sense to have 'allowDiskUse' be opt-out (default=true) rather than opt-in. This would require a number of changes, not just in the server, but also in atlas (for example, on the free tier we'd probably want to continue to disable spilling altogether).
This ticket is meant to track the discussion around switching the default of allowDiskUse.