gpkit: unknown solution status dependent on constraint list ordering

I just ran into some behavior I really don’t understand. These are the constraints for my combustor model, and everything works fine with them.

           constraints = [
                #flow through combustor
                Pt4 == pib * Pt3,   #B.145
                ht4 == Cpc * Tt4,

                TCS([f*hf >= (1-ac)*ht4-(1-ac)*ht3+Cpfuel*f*(Tt4-Ttf)]),

                #making f+1 GP compatible --> needed for convergence
                SignomialEquality(fp1,f+1),

                ht41 == Cpc * Tt41,
                ]

            constraints.extend([
                #compute Tt41...mixing causes a temperature drop
                #had to include Tt4 here to prevent it from being pushed down to zero
                TCS([ht41 <= ((1-ac)*ht4 +ac*ht3+Cpfuel*f*(Tt4-Ttf) + Cpfuel*Ttf*f)/fp1]),

                #comptue the rest of the station 4.1 variables
                SignomialEquality(fp1*u41, (u4a*(1-ac)+f*u4a+ac*uc)),          

                #this is a stagnation relation...need to fix it to not be signomial
                TCS([T41 >= Tt41-.5*(u41**2)/Cpc]),
                #here we assume no pressure loss in mixing so P41=P4a
                Pt41 == P4a*(Tt41/T41)**(1.313/.313),
                #compute station 4a quantities, assumes a gamma value of 1.313 (air @ 1400K)
                u4a == M4a*((1.313*R*Tt4)**.5)/hold4a,
                uc == ruc*u4a,
                P4a == Pt4*hold4a**(-1.313/.313),
                ])

however, these constraints cause the model to return unknown

           constraints = [
                #flow through combustor
                Pt4 == pib * Pt3,   #B.145
                ht4 == Cpc * Tt4,

                #making f+1 GP compatible --> needed for convergence
                SignomialEquality(fp1,f+1),

                ht41 == Cpc * Tt41,
                ]

            constraints.extend([
               TCS([f*hf >= (1-ac)*ht4-(1-ac)*ht3+Cpfuel*f*(Tt4-Ttf)]),
                #compute Tt41...mixing causes a temperature drop
                #had to include Tt4 here to prevent it from being pushed down to zero
                TCS([ht41 <= ((1-ac)*ht4 +ac*ht3+Cpfuel*f*(Tt4-Ttf) + Cpfuel*Ttf*f)/fp1]),

                #comptue the rest of the station 4.1 variables
                SignomialEquality(fp1*u41, (u4a*(1-ac)+f*u4a+ac*uc)),          

                #this is a stagnation relation...need to fix it to not be signomial
                TCS([T41 >= Tt41-.5*(u41**2)/Cpc]),
                #here we assume no pressure loss in mixing so P41=P4a
                Pt41 == P4a*(Tt41/T41)**(1.313/.313),
                #compute station 4a quantities, assumes a gamma value of 1.313 (air @ 1400K)
                u4a == M4a*((1.313*R*Tt4)**.5)/hold4a,
                uc == ruc*u4a,
                P4a == Pt4*hold4a**(-1.313/.313),
                ])

the only difference is that this constraint TCS([f*hf >= (1-ac)*ht4-(1-ac)*ht3+Cpfuel*f*(Tt4-Ttf)]), was moved from the constraints = [] to constraints.extend([])…any idea what’s going on here?