It makes sense that
Hypergeometric2F1[a, b, b, z]
is
(1 - z)^-a
And also, it makes sense that
Hypergeometric2F1[a, -3, c, z]
is at least in most cases a polynomial (except maybe when c = -3):
(2 c + 3 c^2 + c^3 - 6 a z - 9 a c z - 3 a c^2 z + 6 a z^2 +
6 a^2 z^2 + 3 a c z^2 + 3 a^2 c z^2 - 2 a z^3 - 3 a^2 z^3 -
a^3 z^3)/(c (1 + c) (2 + c))
So, the question is then that why
Hypergeometric2F1[a, -3, -3, z]
is the polynomial
1/6 (6 + 6 a z + 3 a z^2 + 3 a^2 z^2 + 2 a z^3 + 3 a^2 z^3 + a^3 z^3)
And not (1 - z)^-a?
Obviously, the limit is different depending on which variable approaches the value. Is this well-defined based on some common knowledge or is it an arbitrary decision which value to use?