Here is what inertia relief does: Let's say that we have a flexible free body (e.g. satellite in space). We apply some forces to it. The forces are going to do two things. First, they are going to cause a rigid body acceleration according to F=ma. i.e. the center of mass of the object is going to accelerate based on the sum of the forces. But, because we have a flexible body, these forces are also going to cause local deformations and stresses. We don't care about the rigid body acceleration of the center of mass. We only care about the local stress.
Now, we could solve a transient dynamic solution (i.e. sol 109 in NASTRAN), to get both the rigid body acceleration and the local deformation, and then afterward subtract off the rigid body acceleration to get just the local deformation that we care about. The problem is that would be computationally expensive. We are spending a lot of CPU resources to compute something that we are just going to through away anyway.
So what we want to do is to solve just a static solution (i.e. sol 101 in NASTRAN). But there is a problem. Because the structure is unconstrained (free), the stiffness matrix is singular. We cannot invert it.
We could add a fake support to the structure somewhere, just for the sake of constraining it to get an invertible stiffness matrix. But the reaction forces at this constraint will cause local deformations and stresses that are not real and will mess up the answer.
So we have a problem, we don't want to solve a transient dynamic problem because it would take too long, but a static solution doesn't work.
So here is what inertia relief does. First, we add a constraint. i.e. we make the structure fixed at the boundary instead of free. Then, we add a distributed body force to the entire structure that exactly balances out the rigid body acceleration. That means that the reaction force at our new dummy constraint is zero. So, the system is now constrained, so the stiffness matrix is non-singular, so now we can invert it. But, the reaction force at the dummy support is zero, so it doesn't have any effect on the local deformations and stresses.
This might help to illustrate: https://www.simutechgroup.com/tips-and-tricks/fea-articles/251-fea-tips-tricks-ansys-static-analysis-inertia-relief
So, having said all that, can you use inertia relief for buckling? If you are trying to model buckling of a free body that is accelerating, then I don't see why not. For example, if you have a satellite in space, and this satellite has some flexible structure with thruster attached to it, and you want to know what thrust it will take to buckle your structure, then inertia relief makes sense. The added fake constraints are only going to constrain out rigid body motion which should not affect the buckling results.
But if you have a constrained structure that is not accelerating, then I don't see how it makes sense. I guess what you are doing is deleting the boundary conditions entirely (because they are hard to model), so fictitiously unconstraining the model, and then using inertia relief to put them back in... Inertia relief is going to try to generate a body force over your entire structure in order to enforce a zero reaction at the fictitiously created constraints. But your structure really does have real constraints, and so there should be a non-zero reaction at the real constraints. So the applied body force will be wrong.
Not quite sure how NASTRAN handles that body force. If it uses it with all of the other loads in generating the differential stiffness matrix, then you definitely get the wrong answer. But if it ignores it when generating the differential stiffness matrix, then you might get the right answer, if the supports that got added back in were equivalent to the ones that got deleted.
So that was a really long winded answer... bottom line, although it might work, unless there is something special about how NASTRAN does this that I'm not aware of, I would not recommend it. Spend some extra time to get the boundaries modeled correctly.