I believe you're thinking of Grover's algorithm, which as you remember offers the ability to search for a value in a function domain in O(sqrt(N)). This is trivially countered just by doubling the key length, a 128-bit key probably couldn't be considered secure (average 2^64 evaluations) but a 256-bit would still average 2^128 and 512-bit 2^256. I vaguely recall there was some speculation that applied de Broglie–Bohm theory, if it was accurate, could allow a mild improvement but even then it only changed it to cube-root, which again could be countered with slightly larger keys (in practice 512-bit would still be plenty even then).
At least for now there don't appear to be any major algorithmic implications for symmetrical encryption from a theoretical future scalable general purpose quantum computer. Granted, there may be some practical historical ones, and in a few cases they could be of significance. If agencies have archived data of interest encrypted with 128-bit in the past, they may gain the ability to gain access to it within the next decade or two. A lot of it might be well past its proverbial sell-by date but not all if those using it were counting on it last much longer then they'd be alive.