Suppose there exists a chain of views between 10 and 20 views in length (i.e. view1 is a view on view2 is a view on ... viewN) (as is the case in views_validation.js). In a sharded cluster, an aggregate against such a chain of views may initially fail with a 'StaleConfig' error (but subsequently succeed on a retry) because it will exceed the maximum of 10 StaleConfig retries for each view that it doesn't have routing information for.

This ticket tracks the work to improve this so that such an aggregate does not have to be retried (once such way is to track views in the sharding catalog, another way is for shards to be aware of the views that they own).