[...The paper is 10 pages and says on page 9, "The simulation with Theia not spinning initially yields an orbiting proto-Moon with a periapse at 4.5 R⊕, well outside the Roche radius of ∼3 R⊕. It has a mass of 0.01 M⊕
0.81 M...On page 4, Table 1 is provided showing a proto-earth with 3 hour day or so.
That also is the view found in the new book, "The Earth Had Two Moons", contrary to what I had thought, but cautioned I could be wrong.
It is the proto-Moon that is assumed to have about a 6 hour orbit.
The rule of thumb for terrestrial planets puts the Roche limit at about 2.5R, so a minimum of 3R is safer for the limit. This is the forming region the author took for discussion in his book.
It seems logical that when an impact occurs the majority of the mass lifted up will fall back down. Less and less material will reach higher and higher altitudes and orbits. Thus, the bulk of the mass that survives in orbit would be just outside the Roche limit, IMO.
[Getting a bit astray of the OP, but interesting, is that this event would have created two strong Lagrange points allowing for Trojans. A large mass could have formed at an L4 or L5 or both. Later, as they moved outward due to tidal action with Earth (first hypothesized by George Darwin in 1879), then the instability would have caused the smaller body to impact the Moon.
But the impact velocity of a Trojan would likely be about that of the slow esc. velocity of the Moon (2.4 kps) and a talps (splat spelled backwards, Belton) event. This gentler class of impact seems to explain the farside's thicker crust and thinner KREEP.]
As computer models advance in the giant impact model, more and more special initial conditions seem needed now to make everything work
Yep, initial conditions are always critical to cosmogony and celestial mechanics.
What is amazing is that there are hypotheses that can match most of what we see. But I don't think anyone suggests we have it nailed down yet.