Normalization jump for resonance mapping on resonance

[Juergen Reuter]

When I used isr, I found that the cross section would jump in the mass of the propagator.
My code:

model = SM
beams = e1, E1 => isr
alias prt = e1:e2:E1:E2
process eemm = e1, E1 => e2, E2 {$restrictions = "1+2 ~ Z"}

sqrts = 91.18
cuts = all E >= 100 MeV [prt]
   and all abs (cos(Theta)) <= 0.99 [prt]
   and all M2 >= (1 GeV)^2 [prt, prt]
integrate (eemm) { iterations = 50:3000:"gw", 25:2000 }
real s18 = integral(eemm)

sqrts = 91.19
cuts = all E >= 100 MeV [prt]
   and all abs (cos(Theta)) <= 0.99 [prt]
   and all M2 >= (1 GeV)^2 [prt, prt]
integrate (eemm) { iterations = 50:3000:"gw", 25:2000 }
real s19 = integral(eemm)

printf "s18 = %f" (s18)
printf "s19 = %f" (s19)

And my output:

|=============================================================================|
| Time estimate for generating 10000 events: 0d:00h:00m:08s
[user variable] s19 = 1.694136387477E+06
s18 = 1469934.510309
s19 = 1694136.387477
| WHIZARD run finished.

Z-lineshape seems discontinuous at MZ.

This indeed has turned out to be bug that has been filed in late December to our development team.