clustered_group_atomicity.js uses waitForIndexBuildToStart which does not guarantee that the side write table has been set up because collection scan hasn't started. We should use waitForIndexBuildToScanCollection so avoid the concurrent write being written in the window before the collection scan starts because the write ends up as part of the collection scan rather than as a side write. Wait for the scan to start ensures we generate the side write to test the smart grouping for clustered collections.