First question (answer):
There is no effect on Horiz. shooting, as you have already corrected the sight adjustment for that particular angle (lets say 0° (horiz.)). This would hold true for any angle of shot, whether uphill or downhill. Shooting at the "exact" same angle for each shot will allow the arrows to "group" similarily, but the group will just be displaced. Don't forget gravities effect on those extreme up or down shots, as they would naturally be "High".
Second question (answer):
No!, If you are a target shooter and always shoot on level ground.
Yes!, If you are a BowHunter, where you may need to shoot at any angle, or any angle that may not be the same as that of which you sighted the bow in at.
Most people here are confusing "Cause" with "Effect".
The "Cause" = A mis-aligned 3rd axis. (the axis perpendicular to the riser face).
If the axis of the bubble level isn't perpendicular to the "Riser's" face, then the bubble level will have one of its endcaps pointing towards, or away from the shooter. This is not a problem shooting on level ground, because reguardless of the endcaps orientation on this horiz. plane, it is still level with the ground, and will read "true level" reguardless how you spin around. The problem is when you need to rotate the bow in a vertical manner to shoot up or down.
Lets say you are a Right-handed shooter, and the your sight bumped a tree, as mentioned earlier in this post. The impact has bent the sight back towards the shooter. The bubble level's left endcap is now closer to the shooter than the right endcap. Now, if you rotate the bow upward for a shot, you can visualize that the left endcap will be now be lower than the right endcap. The shooter see's the bubble displacement to the right, then cants the bow to the right (CW) to compensate. Thus the "Effect"
The "Effect" = Canting. (Canting (with regards to archery) = rotation about the axis of the shooting arm).
The effects of canting can be simulated on level ground.
To simulate this overall effect, and its magnitude on your shot displacement, take your bow and shoot a few arrows on level ground to establish a baseline group, keeping your bubble level reading horizontal. Next, cant your bow slightly, either right or left, like you see the Indian's doing on TV during a buffallo hunt on horseback (how is that for descriptive). Cant the bow until the bubble on your level just touches it farthest extreme. I use this example, because you will need to duplicate this canting angle for each shot if you expect to get the same size group. Don't be affraid to cant your bow 45° or more (like the Indian's), just figure out a way for you to reproduce that cant angle for each shot. Now take a few more shots. Notice that your "group" should have remained the same size, but your "Group" placement on the target has changed.
Typically, if you "cant" the top of the bow to the right (CW), the shot will displace to the right, and a left "cant" (CCW) will displace a shot to the left. In both cases the vertical shot displacement will typically be lower.
This is the same effect you have with canting a rifle, twisting it right (CW) or left (CCW) . The reason for the shot displacements have to do with the horiz. (windage) & vert. (drop) adjustments you have "dailed-in" on the sight.
Image you have "Zero" windage, and +12" of drop adjusted into your sight. Now exaggerate the effects of canting by rotating the bow/gun/sight a full 90° (CW). You now have 12" or more of right windage, and "Zero" for drop. Your shot would land approx 12" to the right and 12" low. You can see by this example, that the further the shot the worst the error, mainly due to the ever increasing trajectory corrections needed in the sight.
Sorry this was so long, but I hope this helps answer all those remaining question on this subject.