Strings are still fundamental in M-theory. While it is true we do not know what the final theory will look like, it is highly unlikely that strings will be completely replaced by something completely different. What we know about M-theory is still based on string perturbation theory and non-perturbative dualities between string theories and also gauge-gravity duality which also involves strings.

The important things that make string theory possible, like modular invariance (responsible for UV finiteness) are all connected to the unique properties of Riemann surfaces, so whatever M-theory is, it cannot completely do away with the 2d worldsheet of strings or something close to it.

How can that be? M-theory lives in (at least) 11 dimensions, while strings are not consistent in 11 dimensions. Strings only make sense in 10 dimensions, and 10-dimensional objects are not fundamental in a theory which fundamentally lives in 11 dimensions.

Are there any two objects for which one can prove that they are not dual to each other? Strings can be dual to fields, theory in one number of dimension can be dual to a theory in another number of dimensions, classical ER can be dual to quantum EPR, ... Perhaps string theory can be dual to LQG? Where is the limit?