Seems very consistent?
Initial rockets are hitting the side of the stair, shielding opponent from the splash.
Follow up rocket it is hitting the top of the stair, and there is direct LOS from the impact point to the target, so they get hit by the splash.
That being said this situation should be fixed. A small edge shouldn't shield the target, and explosions hit go through these tiny blockers.
Correct, this is working as intended as per the current implementation, the splash volume is very simply projected hemispherically. When you fire at the horizontal quad of a stair-step this hemisphere faces up, when you fire at the vertical quad this hemisphere faces you. I've now brought it up as a design change and we are OK with "modernising" this. I imagine we are going to create ramp-shaped envelopes for the stair pieces that any splash-creating projectiles will collide against instead of the normal collision hull complex and that will give you a more intuitively oriented hemispherical projection of the damage without affecting how things currently work when splashing around other convex accidents.
The slight problem with this is that if you are shooting sideways alongside one of these stairs there's now going to be this invisible right triangle that splash projectiles would collide against, but this situation should be rare and we can probably avoid this by making sure our stairs are always bevelled to the sides whenever they are going to be used uncapped. This will not affect how movement physics work with stairs.
devtrackers.gg