I have looked at the code from a 1499 based 'upgraded' PROM supplied by a a well known chipping company and run a compare with the stock 1499. Two bytes were changed. On analysis of the disassembly I found that one of the changed bytes breaks the call to the error subroutine made when the overboost criteria are met (and I mean breaks - the call is broken so a bunch of arbitary code gets executed instead) and the other change is to an unused byte and its purpose is simply to balance the change made to the first so that the checksum comes out unchanged and the PROM passes the power on self test.
In practice this meant that something, to the driver, much like the normal overboost fuel cut occurred on overboost but the CEL did not light nor did the code set. The cause of the abrupt power loss was in fact the 'crash' caused by the arbitray code execution I mentioned above rather than any deliberate engine protection measure.
If you look at the highway.boost.creep.ecu log at 245.8s you can see all the parameters that are put through a moving average smoothing filter get reset to zero (speed for one) and some unsmoothed variables are set to zero as well, a/f ratio & knock count included, when the corrupted overboost call is executed and this is what causes the ECU's confusion and the fuel cut.
I hope they had later versions better than this because its pathetic that this was given to someone to put in a car.
This is my opinion based on my interpretation of a code disassembly. I may have got it wrong, although the evidence in the log I mention above supports my interpretation.

