I was thinking about this in the context of the same-sex marriage debate because state recognition of marriage seems to require a lot of linedrawing. It is widely agreed that the state has to impose limits on the number of people who can get married (2? 3? more?), the age of the couples (with or without parental consent), consanguinity (whether cousins can marry), etc. It can be hard to justify the exact lines that get drawn. For example, in many states, couples can get married at 18 without their parents' consent but 16 with their parents' consent. The exact line here seems pretty arbitrary. Why 18 years, and not 17 years and 323 days? Why 16 years if a parent consents, and not 16 years, 4 months, and 19.5 seconds? No matter who draws the lines required to define marriage, some parts of the definition are going to be rather arbitrary.
My claim — albeit only a very tentative claim, as this isn't my area and I haven't looked closely at the cases — is that the fact that some line needs to be drawn, and the legislature unimaginatively drew it in some relatively traditional way, itself helps provide a rational basis for the legislature's approach. It seems sensible for a state legislature tasked with all the line-drawing of defining marriage to stick with the common answers to the problem. Put another way, deference to preexisting practices in areas that require complicated line-drawing is a sensible default even if we lack a clear argument for why those preexisting practices are normatively preferable to other ways of drawing the lines.
Weak tea, I say. The slick move here is to compare a "line drawing" problem with respect to age (where the range of lines that could be drawn is literally infinite), to that of gender, which is a binary problem and isn't really a "line drawing" problem to begin with so much as it is a yes or no question. In situations like age of consent or even number of partners one can simultaneously marry, the broad range of potential solutions may counsel judicial deference to the seemingly arbitrary line the legislature has picked. It does not follow that this extends to "problems" where a legislature is debating between essentially two choices: allowing same-sex couples to marry, or forbidding it.
One could argue, of course, that the relevant basis for comparison is that we're changing the "definition" of marriage, and the line drawing problem is where we draw the line for subsequent changes. The problem is that if this rationale suffices for rational basis inquiry, it becomes limitless -- any change in law or policy could conceivably raise the question of "how else could we have to change it", and that simply can't be enough to uphold the legislation if the rational basis test is to retain any sort of critical bite.