Cross-input signature aggregation is not a taproot usecase, it is a feature that could be built with Schnorr-style signatures. The Taproot softfork did not include this feature, because research into cross-input signature aggregation was not progressed far enough. Waiting to include it would have delayed the Taproot softfork by years.
Is it at the point where it's secured to be done now? I'd love to learn more about that
reply
It’s still a research stage proposal, and implementing it will now require a new type of outputs, so it might get shipped with another proposal that requires new output types at some point
reply