I don't want to disagree, but I'm going to say it depends on the field. I've met people who have loved their job there, and others who didn't seem to enjoy it. I've worked at two government labs now. It may be a YMMV thing.
In my scenario, as a postdoc with interest in computational science, I would be going into one proposing a project. I've already done this for one, and am scheduled to give a one hour public seminar this summer describing the work to anyone who feels like showing up (but I have not yet decided if I will go longer term with this - but regardless I will do it anyway because I do enjoy giving presentations about my work). This part is very difficult. But 1) if you're not a postdoc, most likely you won't even get the chance to do this, and instead you will be trying to fit into an "available positions" (like searching for a job normally) with specific tasks and goals, and of course 2) not just coming up with ideas, but the whole plan including the approach, execution, timelines, etc. And I can say since some of this was discussed already before I came back to finish up my degree, this is pretty much a done deal if I accept. All of this happened because I had this passion while I was working where the ideas just kept spilling out left and right, and all the directions I want to take the work (but bummer I have to settle just one ). If you're not passionate about your work, then a PhD and postdoc related work is not for you. Because that's where all the creativity, innovation, and most importantly - the drive to pursue comes from. Academia teaches you the other necessities (research approach, writing, oral presentations, etc).
And then the advantage with computational science is that 1) it is huge in government laboratories, 2) there are no shortage of models that need to be updated, improved, or just flat out rebuilt and 3) start up and maintenance cost is very, very little compared to other areas. If you're doing experimental work, again, you probably won't even get the chance to do what I am doing because how insanely expensive experimental type work is, government regulations, etc, compared to computational. I'd say if you're an experimentalist, you probably don't want to look at the government unless they have positions available that really fit you. Both are needed in the end, but if there's a side there is an obvious more demand for - it's the computational one. Because the code you build will work its way into some huge larger package that all of these government labs have.
A government lab to me is academia-lite. The expectations there are ~1 good publication a year from someone like me (not an exact metric, but the "amount of work"), and they are happy. I'm used to a rate of ~3/year right now (which is too much and exhausting especially when you feel like you're not being treated like a professional). That's not to say I will not find a specific job description in the industry, that I will end up interested in and at least apply.
Sometimes I do wonder if people mix up computational science with computer science. Because computational science is, and will be (probably forever) a field in demand. But it really has nothing to do with computer science. It's engineering based.
P.S. My ATOT walls are in no way representative of my technical writing.