The post-doc is now effectively mandatory in computer science academia like other fields. I don't know where this figure of 75k for an astrophysics post-doc comes from. The only 70k+ post-docs I've ever heard of are at government labs like Los Alamos. At Harvard, for example, post-docs make less than 50k per year (I'm a grad student there and have post-doc friends). If Harvard's Smithsonian center for astrophysics cannot pay market demanded prices for post-docs, then I just don't buy this claim that Santa Cruz could.
I second the sentiments mentioned above that the years and low-pay of a post-doc are depressing now that it is overwhelmingly likely that it won't lead to a tenured position.
I wouldn't say it's mandatory; about half of my grad-school friends did post-docs, and half didn't. It depends strongly on your sub-area, and in part on how flexible you are on location and type of school. In some cases, people who take post-docs could've gotten professorships, but not at top-tier research universities, and preferred a year of post-doccing at Stanford to taking a professorship at a low-ranking school.
I second the sentiments mentioned above that the years and low-pay of a post-doc are depressing now that it is overwhelmingly likely that it won't lead to a tenured position.