Soe more failures in longer running tests happened because trunk_file_write_recovery removes objects from the active list before they did manage to get replicated to another node. so back to th drawing board, comments on the concept are still welcome, through.