Redirect: Whether to order the equivalent nodes by ctime to determine the node used as cache source

[unkcpz]

This discussion is redirected from the PR #5197
Currently, if there are equivalent nodes (have the same hash), method _get_same_node which decide the node to be used in caching will return the node when the first match is hit in the DB.
The PR change the behaviour that the nodes will be ordered by ctime and the latest one will be returned.
It is not sure: a) if it is an improvement and necessary functionality. b) will this implementation have a large impact on performance if there are equivalent nodes in a very big DB.

Still, something can be done to decide whether we will have this in the code:

• Find a large DB (millions of nodes) and create a few identical nodes in it.
• Benchmarking the _iter_all_same_nodes with and without order by ctime in a QueryBuilder
• On Django with db_index=True
• On SqlAlchemy with no index on the ctime of node.