I ran several profitability calculators, and even at $85 cost and free electricity, these units don't seem to make any sense.
Even if you consider selling your GPU's, not sure it's a wise strategy to invest cold BTC or Cash to buy Erupters.
By the time they breakeven, they'll be yielding so little profit to even justify the attention to keep them running...and that's assuming they'll survive 10+ months of non-stop mining.
It's more like giving Asicminer an interest-free loan, or betting BTC will be >$500 in next two months, or a combination of both.
You are correct on all counts apart from the last one - if you are betting BTC will be $500+ in the next two months, there are many smarter ways to express that view. (such as, erm, buy BTC today?)
Regarding the question of why anyone would sell a profitable asic - there are some good reasons actually:
Even if you can build thouands of asics, that's not the same as operating thousands of asics 24/7. The logistics of space, power consumption, cooling, security etc scale very fast once you start looking at hundreds or thousands of units.
Selling an asic is instant ROI - operating an asic is ROI over time, which always carries more uncertainty. Asic manufacturers have bills to pay, some cash in hand is always needed.
And finally the 51% issue - even if someone has a highly efficient asic, getting close to 51% of network hashing power would be detrimental to their long position in bitcoin. Selling units would be a way to further increase profits without increasing their share of hash rate.
But realistically, at the moment at least, the biggest reason to sell is because demand > supply, which is reflected in the ridiculous prices you have noticed.