None of the laptop packs I've torn apart have had any cell balancing, and they very often fail due to one pair of cells being significantly out of balance with the rest.
Yes - in most cases, because they implement cell-level measurement using any of the many broken-by-design chips. I have measured these "BMS" devices drawing orders of magnitude too high quiescent currents (hundreds of uA), in itself causing overdischarging of the cells when the user leaves the pack in discharged state for some weeks or months, leading to self-destruction (a li-ion cell itself is OK to be stored fully discharged, even for years, but any external load in this state destroys it easily).
In some cases, instead of complete self-destruction, these laptop BMS boards "just" unbalance the pack due to differences between quiescent currents per cell tap - for example, if one tap consumes 20uA and another 50uA, it's 262 mAh of imbalance (or about 5% of typical 2*2600mAh capacity) in just one year.
It's true that laptop BMS boards do not always provide balancing - balancing would be needed to compensate for the parasitic unbalancer circuit. I have seen many that provide balancing, though. Most balancer circuits, however, have limited operating region where they can do balancing, and "wrong" usage patterns still cause significant imbalance.
The most robust packs I have seen are totally BMS-less power tool packs such as used by Bosch.
Though, some laptops may have an additional imbalancing issue of having a huge local CPU/GPU hotspot, which could heat one cell to significantly different temperature than the others. In these cases, however, capacity loss and ESR rise is often even a bigger issue than imbalance; manually balancing these cells usually remains a partial remedy only.
Li-ion management IC market, lead by the usual big guys, is totally crazy, completely driven by Powerpoint and handwaving instead of scientific study of battery technology. For example, as anybody who has done any analysis knows very well, non-dissipative (redistributive) balancing almost never makes any economical or ecological sense, yet we have a shitload of redistributive balancer ICs, usually using uselessly inefficient topologies such as neighbor-only switched capacitor.