Requerying every 10 minutes would destroy many positive features of the network. That type of activity would cause 90% or more of all query traffic to be requeries and reduce user initiated queries to 10% or less. Given that this would attempt to jam too many queries through the network, all queries would more quickly die off (from flow control) and the overall experience would get worse.

We have download meshes which connect sources for a file together directly. If new sources for the file become known, then these sources will show up in the download mesh and get propagated to new downloaders. This is a low cost way to get new sources other than through requerying.

If your download dies, it requires user intervention to restart. This is the best way to prevent excessive requerying. Any automated requerying is too much.