Lets talk for a moment about mouse movements.
If you're playing on a Window PC, then PokerStars doesn't track the mouse, Windows does. PS can tell Windows to change the cursor, but Windows directly controls where the cursor is on the screen based on interpreting mouse movements coming from the hardware driver.
What Windows does next is, if the mouse is over a PokerStars window, it tells PS, "The mouse has moved over this window, it's here (x,y coordinates) now" in something known as a "MouseMove message" (Windows API documentation:
WM_MOUSEMOVE). This message also tells PS if any certain are clicked. These messages transmit lots of numbers to the PS application.
Now if I were coding an RNG that required a seed, and I wanted to randomize the seed, one thing I could do would be to capture all those mousemove events, along with mouseovers, keypresses, and such, and just smash them all together (concatenate) and use those smushed numbers as my seeds. Mousemove events alone generate dozens of numbers and so you could just chunk them up into say sequences of 16 (picking a nice 2^4 number out of the air) then use that as my seed. When collecting these thousands of numbers from hundreds of players, you can further randomize the seeds by having each client send these numbers back in a constant stream, picking one digit at a time from each player by looping through all the active IP ports in sequence, so that no one player controls more than one digit of any seed, and further you can pull those digits only when you need them but continue to let the data stream, meaning you can never predict when a given players stream will have a given digit picked.
If I had to write an RNG that guaranteed unpredictability, that's how I'd go about the first part - identifying the seed. That would certainly pass any inspection possible by any independent reviewers with half a brain.
But sure, lets pretend like moving the mouse one inch at exactly 278 deg when the clock timer is at 12 will give you AA. That kind of obvious code would never be detected in an independent review.