The internal aggregation stage $_internalAllCollectionStats applies an optimization in case of having a $project stage behind. This optimization may prevent the aggregation from delivering the storageStats field even when it's requested.

This is an example where we don't get the expected output:

> db.getSiblingDB("admin").aggregate([
    { $_internalAllCollectionStats: { stats: { storageStats: {} } } },    
    {$match:{ns:"dba.collB"}},
    {$project:{"storageStats":1}}
])

[ { ns: "dba.collB", storageStats: {} } ]

This is an example where we get the expected output:

> db.getSiblingDB("admin").aggregate([
    { $_internalAllCollectionStats: { stats: { storageStats: {} } } }, 
    {$match:{ns:"dba.collB"}},
    {$project:{"storageStats.count":1}}
])

[ { ns: "dba.collB", storageStats: { count: 1 } } ]

This bug doesn't affect the final user because this is just an internal aggregation stage used by the $shardedDataDistribution aggregation stage. However, it could cause future bugs if we decide to change the $shardedDataDistribution pipeline in the future.