I believe the balancing actually occurs once the green light stays on. The pulsing is the BMS causing the charge current to be cutoff, when a cell gets too high. This will happen as the cell's voltage first hits the cutoff, which sounds like it is set for about 3.75V. At this point, the cells are about 85% full. Eventually, the cells causing the cutoff will get full enough that the current required to keep "filling it" is less than the trip point, which is really the amount at which the shunt resistors can pass without being swamped (50-100mA, I think...). When all the cells are at this point, the green light comes on solid. This is when the shunts for the fuller cells are bypassing the maximum amount (i.e. -- 50-100mA...) and the cell is as full as it can get. This 50-100mA is then passed up the chain to the next cell. If this next cell is not completely full yet, all of the bypass current will go into the cell. Eventually it will get full, and its shunt will bypass the 50-100mA up to the next cell, and so on. Eventually, all the cells are full, and all the shunts are in full bypass. With such a low bypass current, it can take literally 24 hours in order for each cell to get completely full, if you have cells that are significantly out-of-balance in relation to the rest.
The BMS that Richard and I are doing operates in a similar fashion, but instead of a slow "pulsing", we have logic that keeps the charge current just below the threshold of what would swamp the shunts. The other difference is that our shunt circuits will bypass 5-10 times as much current (up to 1/2A...), so it doesn't take near as much time for the low cells to catch up, and get completely full. Finally, we now have a feature that completely cuts off the charge current when all the cells are as full as they are going to get. When this happens, our LED turns green.
RC-type balancers do, in fact, work by draining down the high cells, while waiting for the rest to catch up, but I do not believe any of the typical imported BMS designs do this. I believe they all work the way I've described above.
One other point is that there are two conditions where a cell, or cells, are not at the same level as the rest. In the first, it is typically just that the cell(s) simply didn't start out as full as the others, but are otherwise as healthy, and still have the same capacity as the rest of the cells. Once they are fully charged, it is reasonable to assume they will stay fairly close to the rest of the cells. I think there are many examples of this that people have reported. The second case is when a cell becomes "stressed", maybe by being slightly over-discharged, and it loses some of its capacity (typically 5-10%...). In this case, these cells will get "full" quicker than the rest. With our BMS, you would see the orange LED for that cell come on a lot sooner than the other LEDs.
-- Gary