I've read both.  As it happens, I had seen them both before, or seen
some of the material in them in your other papers.

So as it happens, you had got your wish - I had at least looked at the
research you've done on manipulation before ... well, "reject it"
isn't the right phrase.  I'm not disputing your conclusion, I don't
believe it's an answer to my argument.

I made a specific argument, to which the counterargument raised by
both papers does not seem to apply.  The situation is not simply the
presence of manipulators.  It includes a proposer who colludes with
the manipulators.

Perhaps it will be clearer if I explain it in your own terms.  In
terms of your analysis, a misbehaving proposer can make the
non-colluding traders' variance of asset clue encompass the entire
possible range and can make the cost of reducing that variance
arbitrarily high.  Feel free to plug c(S[i]) = oo into your equations
in biashelp.  It just implies nobody who becomes an informed trader
can expect a profit, so there are zero informed traders, so the
optimal actions are indeterminate, after which the math ceases to
really say anything.  I hope you can easily see the problem now.

Tom