No, the long line is the longest possible line :). Or rather, you can define a long line $L_\lambda$ for $\lambda>\omega_1$, but it will lack one of the most important properties of the long line itself: homogeneity, or more specifically the fact that the (classical) long line "looks locally like $\mathbb{R}$."
The issue is uncountable cofinality. The ordinal $\omega_1$ has the property that no countable sequence of countable ordinals has it as a limit (put another way: the countable union of countable sets is countable). This means that if we define the long line $L_\lambda$ of length $\lambda$ for $\lambda>\omega_1$ ($L_\lambda$ consists of $[0, 1)\times\lambda\times 2$ with the appropriate (non-product) topology), then no open neighborhood of the point $p=(0, \omega_1, 1)\in L_\lambda$ is homeomorphic to any open subset of $\mathbb{R}^1$, hence $L_\lambda$ is not a manifold (even if we drop the separability requirement). More snappily, uncountable cardinals occurring inside a "longer line" break first countability, which is a fundamental property of $\mathbb{R}$.
This is interesting, because it shows that the real obstacle isn't cardinality, but cofinality. A big theme in modern set theory is that cofinality is in many ways more fundamental than cardinality. (Note that there are uncountable ordinals of countable cofinality - $\omega_1+\omega$ is a silly example, and a more interesting one would be the cardinal $\omega_\omega$.)
