Question: Why is Determinant often incorrect for matrices with polynomial elements

When I use the Determinant function on a matrix with (single variable) polynomial entries with real coefficients I often get an incorrect answer. I know the answers are incorrect because they have a higher degree or a lower lowest degree than is possible given the matrix elements.

However, when I replace the coefficients in the polynomials with rational numbers or I put in the option method=minor, I get the correct answer.

The problem seems to be roundoff error. However, the important error is not simply small changes in the resulting polynomial. The important error is the presence of entirely incorrect powers of the variable and not with very small coefficients.

How does this happen and why does the help page for Determinant( ) not warn of this behavior? In particuiar, why does the help page not say that using Gaussian elimination (i.e., the default) will often give incorrect answers in such cases, but using method=minor will work? Is this behavior known? I cannot find any reference to it on the internet.