Cluster (spacecraft)
en.wikipedia.org
en.wikipedia.org
The $370 million was probably the cost of the four satellites and the rocket + operations. It might include R&D but not most of the scientific operations part of the grant (there is tons of science to do before and after a scientific mission like this).
I don't want to thoroughly engage all the problems with your comment, but I'll point out two: (1) Launching two spacecraft is not a cure for poor system reliability. Systems fail after launch for all kinds of reasons; a 1-in-2 launch success rate and 2 launches does not guarantee a successful mission. You can't build systems that are half reliable and half not. (2) Some things really have to work on the first try: Earth and sun-observing systems that need data continuity with failing earlier systems, planetary missions with narrow launch windows, systems with standing armies of on-ground analysis personnel who can't be put on hold while a replacement is built and launched.
2) I fully agree, and there are many missions that are just so important that we should strive for success on the first try (MSL, some extremely better telescope like NuStar, and probably the satellites in this article), but NASA's/ESA's budget is much bigger than just those missions and I would argue many of them could have afforded second launches without drastically compromising reliability (i.e. without becoming half reliable and half not as you say)
Many of these are monitoring missions with data continuity constraints. For example, solar irradiance monitors are hard to calibrate, so missions need to overlap.
Also, they have a set of scientists that do calibration, who can't be told to go away and do something else for two years while a replacement is built. This kind of happened with OCO, and it was a bummer.
I just read up on the OCO and that really does suck. In this case I doubt $280 mil would have paid for 2 launches without making it half-reliable but perhaps a spread on risk across multiple similar missions would work. If (and this is a big if) there are enough missions that need roughly the same payload volume/size, orbit, etc. they can split the cost of a third rocket and both build a carbon copy backup, or if a backup would be too expensive invest in maintaining the staff and tooling for the most expensive instrumentation for a faster, cheaper backup build later on. If one of the mission fails, they have a third backup launch and a realistic shot of having/making a second satellite. If they both succeed, the launch is sold for profit to someone who really really needs the launch window or donated to civilian and less critical mission satellites.
This is of course stretching into changing NASA bureaucracy and would have tons of unforeseen consequences but would have the interesting side-effect of providing more low-cost or even free piggy back launch opportunities, especially for CubeSats. Also, talking to an engineer at Orbital, the increased volume could really help lower costs for everyone Also, in the case of critical staff for calibration, operation, etc., how the cost/opportunity cost of having them just sit there while the mission rebuilds versus losing them to other projects?
My thought isn't to passively set aside part of the budget for backups but to actively use the resources to create a smart risk management system that (and I'm making a big assumption) looks at the satellite and launch hardware as far more expendable than it is now, in exchange for more overall throughput. Thinking about the OCO, it's difficult to tell if it's an example of why it wouldn't work or if the delay in what (I would think) would be a critical data source on our impact on our planet is an argument for a shift in philosophy. It would be like going from "Once in a while, the launch will fail but it's rare so let's focus on making the satellite stay up as long as possible" to "Rockets blow up all the time so let's make sure we're not the ones stuck without a backup for what is already a ten to hundred million range dollar project." With the rapid progression of everything from electronics to materials to information technology, it might be beneficial to start thinking on shorter mission time lines for Earth orbiting missions.
Rereading your comments though, it seems that I am under the delusion that a larger portion of NASA's mission are exploratory and more flexible versus missions that support critical infrastructure or carry instruments for data that is far too valuable. Or maybe the bureaucracy carries more overhead than I think it does and the days of NASA taking risks like the explorers of the Old and New world are behind it.
A starting point is http://nmp.nasa.gov/ which is a series of missions that specifically accept risk to develop selected new technologies. There are also opportunities for higher-risk science-investigations (http://science.nasa.gov/about-us/smd-programs/earth-system-s...). There is also a pathway through airborne systems, to develop new mission concepts (e.g., http://lidar.jpl.nasa.gov/co2las.html)
It's a long road to space-qualify a new technology, and because system failure is a logical OR of subsystem failures, you can't put too much new tech or new science into a mission. You'll try a bunch of stuff and one thing will fail, and you won't learn much.
Multiple launches of unreliable systems are not a particularly good way to reduce risk. That's not doing engineering, that's throwing darts.
EDIT: I'm sure that the launch provider has some naming scheme referencing the rocket itself, but usage of that is almost universally confined to internal communications. Human spaceflight is perhaps the exception (STS-125, not Hubble Servicing Mission #4).