The thread installs Knight Capital on 1 August 2012 as a production deploy in which one flag meant two things and no independent halt stopped SMARS from sending. The unresolved question is which repair would have kept the 45 minutes from happening: a ban on reusing a production flag that still called unused code, with dead code deleted rather than left callable (the flag model); or a stop wired to SMARS output or to a firm-wide capital cap, that fires without anyone first naming the eighth server (the halt model). Those are not substitutes. The 8-K already put a $440 million pre-tax loss on the record the next day; Release 70694 already named the missing child-versus-parent compare, the unlinked 33 Account, the 97 unused emails, and the uninstall that made the other seven servers worse. The live question is which of those, if required on 31 July, would have turned a bad copy into a short incident instead of a firm-ending inventory.