If balancing is not used at all, then it is quite important to initially balance the cells. Large scale manufacturers can possibly trust whoever is supplying them cells that they are initially balanced, but at least checking is a great idea. If you are a hobbyist or small producer of products, just go through initial balancing. Simple way to do that is to connect all cells parallel first (before this, make sure voltage differences are within tens of millivolts to avoid large currents between the cells), then fully charge to the final voltage.
The job of in-application balancing is to keep available pack capacity maximized as cells age. The safety functions are provided by cell-level low-voltage and high-voltage checks which are orthogonal to balancing; balancing is optional, cell monitoring "mandatory" part of a BMS.
If even cell-level voltage monitoring is not used at all (i.e., no BMS), then it is hugely important to
1) initially balance the cells,
2) make sure cells have similar capacity (i.e. tight manufacturing tolerances)
3) make sure cells are thermally coupled i.e. stressed similarly during the years of use
4) make sure the overall manufacturing quality of the cells is top-notch, i.e., small and similar self-discharge, similar aging (capacity fade, ESR increase) characteristics.
5) quite obviously, connect nothing external to the middle taps
Such no-bms approach in series stacks, even 6s, has been successfully used by companies like Bosch, even though this is hard to believe to many armchair internet experts. That being said, I don't think anyone does this anymore: design or try to find a proper BMS. I have mixed feelings between recommending a no-bms or crap-bms solution - neither is really a good idea, but indeed, a well thought out no-bms solution is certainly better than worst case crap BMS.