I actually do think I’ve invented a data structure for interleaving multiple streams in a single file in a really efficient way. I can’t find something that looks like it, and it seems obviously faster and more logical than anything else I’ve seen. It’s the level below B-trees so not optimised for least amount of lookups within a record table, nor would it be efficient for adding records in a middle of the stream, but for treating data streams like files (append, shrink) I do think it’s significantly more efficient than what else I can find. One of these days I really should submit a paper and see what reviewers say.