Isn't Fisher's exact test an example of sampling in which all margins are fixed in the 2x2 contingency table?

The question below and the answer to them make me confused.

Q: In the 2x2 contingency table, is there a sampling distribution where all margin values are fixed? If so, please explain it with an example.

A: There is not a sampling distribution where all margin values are fixed. This is because in order for all margin values to be fixed, the values in each cell of the contingency table would have to be fixed. However, this is not possible since the values in each cell are determined by the values in the other cells. For example, if the value in the first cell is fixed, then the value in the second cell would have to be fixed as well. However, the value in the second cell is determined by the value in the first cell, so it is not possible to fix both values. Therefore, it is not possible to have a sampling distribution where all margin values are fixed.