Hmm, this feels like it breaks the underlying kad... we have not stored the data... The copies in cache are held because of this no? Until we have a node with sufficient capacity? So really we'd want that extra capcity, or to prune... (And in the end put pressure to pay nodes more to get that space). I think for now this may be an expected issue if we're not growing the network