Deployment
There are multiple paths that could theoretically be used to introduce CISA to Bitcoin. It has been pointed out here and here that a new Segwit version may be the safest option, but more research is welcome and needed before a proposal can be formalized.
Soft fork requirement and security considerations
Activation of CISA, no matter in which form, requires a soft fork since it adds a new verification algorithm. This has serious security implications. Bitcoin can only be as secure as the weakest of its verification algorithms and should spending with aggregate signatures turn out to be vulnerable, this would also put all funds that can be spent with Schnorr signatures at risk even if Schnorr signatures themselves remain secure.
Related Proposals
A few proposals for Bitcoin consensus changes make use of signature aggregation. In the spirit of Taproot, which included multiple changes/BIPs that could have been deployed separately, it is interesting to keep an eye on these proposals and investigate if it may be interesting to integrate them into a wider proposal.
Graftroot
Proposes the possibility to delegate spending of an output to a different script, called a surrogate script. As these surrogate scripts can be chained, the signatures for each of the delegations can be aggregated to keep on-chain overhead minimal.
Entroot
Improvement of Graftroot, combining it with the idea of G’root.
BLS Signatures and OWAS (One-way aggregate signatures)
BLS signatures use a different hashing algorithm that hashes directly to a elliptic curve and they use a curve that allows curve pairing, combining two points results in a number. These properties together allow for non-interactive signature aggregation with the resulting signature still taking only 33 bytes of space. Thus BLS signatures would be perfect for highly effective block-wide aggregation. However, the underlying security assumptions are different that what we have in Bitcoin currently and the cryptography is still comparatively novel.
OWAS is a concept for arriving at one signature per block by using BLS signatures described in a paper from 2013 by an anonymous author.