Assuming 2+ shards

Repro Steps:

sh.enableSharding("foo");
sh.shardCollection("foo.bar", {x:1});
for(i=0;i<500000;i++){db.getSiblingDB("foo").bar.insert({x:i});}
//For Migrations
sleep(60);
db.getSiblingDB("foo").bar.drop();
db.getSiblingDB("foo").dropDatabase();

Check and confirm on the shard servers the amount of data remaining.

Primary shard will have no data for this database.

Secondary shards will still have db files existing as the MongoS did not route the drop command to these shards.

The drop of the database should be routed to the non-primary shards to cleanup. We should do some preflight checks before the drop and then use that to help determine where the drops are needed.