If we have a whole time domain from $[0,T]$ can we have a second order ODE describing $[0,T/2]$ and then a first order ODE describing $[T/2,0]$ ensuring that the solutions of the ODE match up at $0$ and $T/2?$
My initial thought was probably not, as this would mean two states mapping into one state on the boundary, but I am lost on this one.
The solution to the 2nd order ODE on the first time interval may, as you note, be incompatible with the first order ODE. That is, the pair $ (f(T/2^-), f'(T/2^-)) $ may not satisfy the first order ODE. However, if you're willing to allow for $ f'(x) $ to have a discontinuity at $T/2$ then this is fine.