Let us have an two objects (in a box) both of mass $m_0$ far from each other.
They attract each other and at some moment their distance is close and they have substantial kinetic energy. Something was answered here - that we should include potential energy to keep the energy balance close = far.
$2m_0 c^2 + T + U = 2m_0 c^2, \quad where \quad U=-T$
Or I better write: $2 \gamma m_0 c^2 +U = 2m_0 c^2$
Now the same, let's have three states: far = close = bound state: a state where the kinetic energy is removed from the system.
$2 m_0 = 2\gamma m_0 + U/c^2 = 2m_0 + U/c^2 + E_{out}/c^2 $
What happens with the mass of the system?
Far - it should be simply sum of both $m_0$.
Close - I can clearly see (as an observer in rest) kinetic energy ($\gamma m_0$), which increases the mass of the system.
Bound - for simplicity - the two balls stop - surface to surface, I remove $E_{out}$ and I my system weights $2m_0$ minus binding energy ($U$ is negative).
Supposing it was correctly formulated - While I am ok with far&bound - What mechanism in $U$ compensates the system weight when you come close? Imagine two neutron stars in gravitational field or two charged particles in Coulomb field or even some short range interaction...?
Reaction on comment user7027: Thanks. Generating momentum should go at expense of mass, now I see a kind of GTR answer, as far as I understand $\lambda$ has a sense of mass loss/binding energy (per mass or so). So what is the answer for other (non gravity) interactions?