For example, if you can assure us that no more than 30 TH/s will ship per week, the early buyers can be sure to recover their investment in a reasonable time before the mining difficulty equalizes at 20x what it is now.
This is a little silly. At the present difficulty, these units pay for themselves in about 7 days. Even at 20x the difficulty, it's only 5-10 months (depending on when the reward halving happens). Current hardware has a payoff of 7 months. What you're asking for here is for a special advantage for the early receivers. Just release the hardware as it comes available (with the initial grouping that they indicated).
even if BFL could have capacity to rollout 200TH/s weekly, i think it would take 1-3 years for the network to grow to that level.
The current network hashrate (according to bitcoinwatch) is about 12 TH/s. 20x that is 240, so within 2 weeks the difficulty would be at that level. This plus the reward halving is what has killed profitability for video cards and FPGA (outside of BFL which takes trade ins for ASIC). This isn't will kill, if you are thinking about ordering anything other than BFL FPGA right now, don't bother because you won't get your money back. If you want a nice video card for gaming, then buy one and mine to get some of the money back, but expect that by December you won't get the cost of electricity.
Why is it so hard to believe they've already been working on this for months already now?
