Index nodes are being downloaded more than necessary

[inetic]

This happens when two snapshots are being downloaded concurrently. Imagine having two snapshots S1 and S2, where S1 happened before S2 (but I think a similar issue may happen if they are concurrent).

Say there is an inner node N that is shared among the two. Further assume that in both the block_presence field is the same (e.g. Full).

If the client finishes downloading S1 while it has not yet seen N as part of downloading S2. A database trigger deletes N from the received_inner_nodes table and it shall have to be re-downloaded.