I personally would prefer one that ran concurrently with the charging process.
My reason is to prevent ongoing overcharge of the cell(s) with the highest charge acceptance.

Since the cells are truly A123 I am less concerned as I feel in RC we are far more likely to replace them due to the 'time factor' than due to an actual degradation in performance.
1000 cycles with a 100% depth of discharge is a ___ L-O-N-G ____ time. That is "spec" for A123.

Chances are that you will phase them out around the three year mark regardless of the performance level indicated. Even though there may be plenty of cycles left, I expect to do the same.
RCr's are quite conditioned to the NiCad way of life and the penalty for trying for one cycle too many is simply too harsh to accept.
At three full years I consider "value received".