It's a matter of economics and opinion.
First, if 17 have failed already, and the pack is eight years old, there are a few reasons to replace the whole pack.
- The remaining cells have probably deteriorated with age and have reduced capacity compared to new ones, so there would be an imbalance on charge and discharge.
- The remaining cells are likely to fail sooner than new ones.
- Newer cells might use better technology, such as improved sealing.
On the other hand, a new pack is a major expense. If the battery is needed only a short while longer (perhaps it would be replaced by different technology), or if the budget is particularly tight, you might replace the cells, knowing failure is likely sooner rather than later.
As Marcus Miller states, protection is needed, in particular for overcharging and deep discharge. Though Ni-Cd technology has some intrinsic protection, in that at a slow charge rate, gases released by electrolysis recombine, a fast charge will burst the vent and the cell will dry out. With cells in series, this is more critical, since a few weaker cells may be fully charged before others, so the fast-charge current would be going through them. If there are many cells in series, another issue is that on deep discharge, some with less capacity will be reverse charged, which causes damage both through venting and through growth of dendrites, shorting the cell.
- The charge current should be limited to prevent venting, if possible to not much more than the trickle rate, C/10.
- There should be some cut-out that disconnects the pack from load while some charge remains, perhaps 20%, to prevent reverse-current damage.