The Fisher information matrix becomes singular, leading to non-convergence.

The likelihood function does not have a unique maximum; it approaches infinity as $\beta_1 \to \infty$.

The Wald test becomes too conservative.

The residuals exhibit extreme heteroscedasticity.