How hot is too hot for my macbook?

90˚c is typical for that system under high load; anything > than 95˚c is when you might want to be concerned. If the CPU gets too hot, or starts to overheat, the system (typically) will clock down the CPU to prevent overheating. The fans shouldn't wear out, so I don't really see any point in trying to replace what's not broken. Intel Mac's run hot, and the fact that the only fan is an output fan doesn't help with heat dissipation.

Something you could do to help alleviate some of the heat is to download smcFanControl (free) and adjust your fans RPM. Instead of the standard 1800 RPM set it at 3500-3600 RPM, which should help cool down the temp possibly 2-10˚c. You might notice more noise from the higher RPM, although you shouldn't worry about it — MacBooks fans generally max out at around 6,000 RPM, and I've cranked one up to 8,700 RPM before.

After 3 years of walltime and hundreds of years of cpu time, I'll respond to this question to update for anyone interested in modern day values.

My current Macbook Pro (Retina, 13-inch, Early 2015 - 2.7 GHz Intel Core i5, 16 GB 1867 MHz DDR3 RAM) idles around 50-65 C and can get up to 100+ when I run heavy processes in Matlab. Sometimes when I have runaway processes and need to restart, it will get up to 80-90 without doing a lot of work.

This is all pretty normal in my book and cleaning out the fan won't do much. If it is idling above 90, that's a bit different. However, I will say that putting in hours and hours of runtime on a CPU running above 100 C is not the best idea. At the time I was doing this, my Macbook was really my only means of programming on the go so I had to make do. Nowadays, if I ever notice my Macbook getting above 95 due to computationally intensive work, I will throw the job on a desktop.

For comparison's sake - I also have a Linux desktop (intel i7-4790K at 4.0 GHz, 32 GB 1600 MHz RAM) and struggle to get it above 70 C even when running parallel jobs on all 4 cores for hours on end and it typically idles at about 29-32 C.