We can also apply this for grafts, and copy over custom statistics in addition to the indexes.

Here's my take on this:

• In general, we should defer creating BTree attribute indexes on all subgraphs until the subgraph is synced. What we really want is to kick off creating those indexes so that they are ready when the subgraph is synced, but not much before then, but as a first step it's ok to create them when the subgraph becomes synced.
• All other indexes (not BTree or not an attribute index) should be created when the subgraph is first deployed as creating them later can be very very slow with bad consequences for the rest of the system.
• For copies and grafts, we do not want to create the indexes that we create by default for new subgraphs. Instead, we should look at the source of the graft/copy, and recreate all the indexes that actually exist on the source