Maybe
Argument for yes:
Polearm Gamble:
Benefit: When a nonadjacent enemy enters a square adjacent to you, you can make an opportunity attack with a polearm against that enemy, but you grant combat advantage to that enemy until the end of the enemy’s turn.
It does not, like other powers, indicate "on their turn." Furthermore, the natural impossibility of taking opportunity actions on your own turn prevents this power from triggering from your own forced movement.
Furthermore, the enemy must be non-adjacent on the move (defined as leaving 1 square for another, not the move total), so if the enemy starts adjacent to you, a slide of 2 squares is necessary for Polearm gamble to trigger.
Argument for No:
Here:
The rules say no OA's are provoked by forced movement, and this was meant to include PG and/or it would be an unintended side effect to allow this to function like this.
Yes, both those pushes are valid.
When you push a creature, each square you move it must place it farther away from you. [RC 211)
Now, 4e measurements use taxicab geometry, or Chebyshev distance, so the effect is... often nonintuitive. I'm pretty sure 4e floors are constantly shifting hyperplanes. My players have developed the mantra, "Circles are squares, squares are circles, and never draw a triangle on the grid!"
How to tell if your push is valid:
To tell if a creature is moving closer, further, or the same distance, count the number of squares the creature must enter to be adjacent to you if he starts in the first square, or the second. The more squares he must traverse to reach you, the further he is from you.
You also need to have line of effect to every square you push him into. He can't enter blocking terrain, and ignores difficult terrain.
The flavor text describing a power's visual appearance has no effect on the power's mechanical application.
Weird.
This actually means that the squares you can legally push a creature into differ, depending on if the creature is head-on with your square or adjacent by corners, because any square adjacent to both you and him is invalid for a push when he is also adjacent to you.
However, so long as you fulfill the above prerequisites, you're free to bounce the guy around zig-zag-style; nothing says the movement needs to describe a straight line (and given 4e geometry, I'm not sure what a straight line would look like anyway).
For completion's sake, pulls and slides:
For pulls, each square must be closer to you than the last, and slides don't care what your position is related to the target.
Best Answer
There's a slightly-complicated chain of rules, but this actually is answered by the rules, at least in a mechanical way.
Both a successful Grappple and a Grab apply the Grabbed Condition to a character. Grabbed, in turn, applies the Immobilized condition, which says, in part -
So, if a creature is held in place by a Grab or Grapple, any force that tried to move the creature out of the square would need to make a check to do so. Moving the target does not automatically break the effect, but I would rule similarly to older editions, where if the target moves out of the reach of the grabber, the grab is broken (the arm, mouth or tentacle has a limit to how far it can stretch). I would also rule similarly if applying forced movement to the grabber.
As far as what sort of check would be required, if a character is physically heaving another character out of the way, I'd allow the acting character to use Athletics as an Escape check for the trapped character, although that isn't an option explicitly allowed by the rules.