While renormalization of the state vector proved to be necessary in “normal cases”, it seems to cause problems with systems that use non-unitary interaction picture, but do not inherit from Exact (Cf. PumpedLossyMode_NoExact)
The understanding at this time is that this is because in such cases the state vector of the system is actually not normalized. (It would be an error physically – or even algebraically – to take it normalized.)
Ticket moved from /p/cppqed/bugs/2/
Issue closely connected with http://sourceforge.net/p/cppqed/feature-requests/34/