Its not that hard to explain. First tier ISPs buy their bandwidth in kilobits/second not kilobits, so, used continuously, a 1 Kb/s line can download 2.592 Gb/month (324 MB). If I've got 100 customers with 30 b/s lines, each (soft) capped at 2 MB, they will only reach 200 MB, so I've got 124 MB unused at the end of the month.
But I can't sign up more customers because then, at peak times, they would be competing for the same 1 Kb/s line (and only getting 10 b/s - if they all use it at once).
So I provisionally set their (hard) cap to say 3 MB each and shape their usage. The line will be used more and they all benefit. But now I sign up another customer who's 2 MB cap needs to be guaranteed, and suddenly the hard cap must drop slightly. (I need to shape their usage to encourage usage (for less time sensitive activities - P2P) at not peak times.)
Similarly, the theoretical maximum hard cap is a function of the time remaining and the line speed. And because the line isn't used continuously at its maximum throughput this theoretical number drops as the month goes on. I can't hard cap at 3.24 MB on the first of the month and steadily decrease this number.
So I set a provisional hard cap that has been statistically calculated to encourage maximum usage of my 1 Kb/s line, (ie: encourage use at off peak times, and discourage cramming at the end of the month) while still maintaining a high probability of each users speed at peak times being acceptable to them (say 95% of their quoted maximum). This cap can obviously change, due to it being a number based on certain probibilities (established through prior usage stats). the fact that the users know that this cap can change, and can't be guaranteed, also feeds the statistical model.
I think that after several years (I never called them smart) Telkom has finally realised that their statistical model is accurate, and that they can disclose the current estimated hard cap to their users. I agree with this as it allows more efficient use of my cap. I also think that there's nothing stopping other ISPs from doing the same thing. IMHO Telkom is being very competitive in a competitive market (the barriers to entry are relatively low). It's only anti-competitive if they're getting their wholesale bandwidth cheaper than others.
The question we should really be asking is: how can we bring down the cost of bandwidth, so that the price/gig drops significantly? We know the answer to this already, and plans are afoot to make this happen (Seacom et al). I think a truly competitive ISP will drop its prices by 90% two months before Seacom lands, and sign up EVERYONE, but we can only wait and see (and demand).
Off topic: is kilobit kb or Kb, Google spellchecker says Kb, but I've always thought it's kb.