3D software I think has long used the principle that, what you're dragging, is the surface of a sphere centered on the origin (of the coordinate system, or the work, or of a fixed distance in front of the camera), that sphere being tangent to the view plane. So, left-right is a vertical-axis rotation (yaw), up-down is a horizontal axis rotation (pitch), and there's no way really to do a depth-wise rotation (roll) directly, which shouldn't be surprising with just two dimensions of control (a modifier key would do fine here, or assigning drag regions at the edge or corner, say; I forget if I've seen such methods, probably?).
I've never used any that had the... just strange movement that Altium does. I have no idea what it's modeled after, if anything. I've never thought about it hard enough to figure out exactly what it is, mapping SHIFT coordinate and mouse coordinates to actual rotation. Like, for axis-aligned moves (relative to the SHIFT coordinate), it's like it's on a projected sphere, but as soon as you go off axis, it just flips around, all...weird. There's a locus of, if not a singularity, then very rapid rotation along some axis, whose angle is related to the line between clicks, or something.
It's just so... bizarre. I recall everything from ancient Java applets to Solidworks getting it "right". Like, it's the default go-to rotation to do it like that [projected sphere], yet Altium went out of their way to do something else. Y'know?
*In older versions. It was updated in, uh, AD18 or 20?. It's just different enough that I always second guess myself which way I want to move...
Mind, I hardly claim to have made an exhaustive survey of 3D software. And there are other standard methods, I think which apply more in specific cases -- like game editors may prefer views typical in-game, e.g. first-person (rotation centered on the camera, and usually cylindrical i.e. left-right rotates around a fixed vertical axis, rotation not cumulative**).
So, it's just one of the myriad oddities I've gotten used to (heh, and honestly one of the few I can remember and articulate, beside all the others that I don't even think about how weird they are anymore..) over the years using AD.
(**Note that an X then Y rotation is different from Y then X, indeed this is the usual way to correct for roll when a direct control isn't available. Strictly speaking, it's completely controllable, there's no need for a third axis; but it may be convenient to provide one.)
Tim