21:34
Basterra-Mandell consider $E_n$-algebras? I was definitely not aware of that. You seem to be pointing out that there is no proper functoriality for factorization homology in the algebra variable. Functoriality for extended TQFTs in general is a bit tricky, since there are no non-invertible natural transformations between them.
So you have to consider either TQFTs with singularities or lax natural transformations (I'm not sure what the expected relationship between these is) to get a notion of non-invertible morphism between TQFTs. But a bimodule between $E_n$-algebras should then determine such a 1-morphism, etc.
Not that I'm claiming anything like this would make it a "cohomology theory".