My own understanding of Grover is not as solid as I would like for it to be, but I believe that as long as a suitable quantum circuit is built, it would allow brute-force guessing of the master password itself (not necessarily the 256-bit encryption key). In this case, the effective master password entropy would be cut in half, which could be a real problem.
However, it is not a given that quantum cryptography will significantly weaken our current master passwords, even if Grover’s algorithm can be brought to bear. The theory only proves that if the classical search space requires 2N guesses for cracking a password, then a quantum search will reduce the number of required guesses to 2N/2. However, the time to crack depends on the performance of the hardware used, so the classical cracking time (TC) and the post-quantum cracking time (TPQ) would be given by
TC = AC·2N
and
TPQ = APQ·2N/2
where AC and APQ are coefficients that represent the effective hashing rate of a classical and quantum computer respectively (i.e., the time it takes to test a single guess); N is the master password entropy.
Therefore, HNDL attacks will be relevant for AES-protected keys if and only if
APQ « AC·2N/2
Thus, I think the jury is still out on whether this is a realistic threat or not.