Dana DeLouis wrote...
If anyone is interested. *This problem should converge to a solution
quickly even if we (or Excel) are not accurate in our derivative. *I
believe there is a logic bug somewhere in Excel's IRR algorithm.
....

Given the problem with this particular use of IRR (seems to fail due
to larger than usual values), my money would be on an unfortunate use
of Single rather than Double somewhere along the way. It's hard to
believe there'd be overflow. I suppose the bug could come from
bracketing the range in which the IRR lies.