They go to very long lengths to avoid anything that would have the shuttle detach from the 747 in the first place.
Back in the day (get off my lawn!) I flew an escort mission for the 747 with the shuttle attached. This was in the 90s and the shuttle had landed in California and was being transported back to Florida for the next launch.
When the 747 has the shuttle attached it can only climb to about 25,000 feet - much lower than the mid-30s that a normal airliner cruises at. The combined aircraft also are not stressed for anything greater than moderate turbulence. So for the escort mission we took off in a C-141 with a NASA guy in the jumpseat. The 747/shuttle combo took off about 15 minutes behind us. We flew the same altitude and route that was planned for the 747 and reported flight conditions back to them. If we encountered any unplanned turbulence, the 747 would deviate around it or turn back, depending on how bad it was.
Overall one of the most enjoyable things I did while flying the C-141. The NASA 747 pilots are great guys, and very good at what they do. So, there are no contingency plans for landing the shuttle if it becomes detached from the 747. The contingency is to fly the 747 in places that won’t press the limits of the attachment points.
I was at the Museum of Flight at Boeing Field when their company chase plane (an F-86 in Boeing colors) took off. About 30 minutes later it came back escorting an E-6 (I think, U.S. Navy derivative of the 707) with the top 1/3 of its vertical stabilizer missing. I heard later that they were doing flutter testing.
I assume it failed the test. Passed the landing, though.
Although “simply” losing the rudder is a survivable accident (AT961).
Losing all hydraulics/loss of control of control surfaces but having the rudder/stabilizer still intact is also survivable (UA232).
What made JA123 particularly nasty was the combination of both: loss of the rudder and vertical stabilizer in conjunction with complete loss of all hydraulics so the plane had no means of controlling the additional control surfaces, so even with differential thrust they couldn’t control the plane.
Not that this has anything to do with the OP’s question, it’s just an interesting aside and I geek out on this stuff.
By putting a person aboard you’ve pretty much guaranteed at least one fatality. Whereas with anything resembling sensible course selection, the empty Orbiter could be transported so that in the case of a separation the number of fatalities should average much less than one.