Based on my benchmarks, the SanDisk Extreme is as much as 2-3 times as fast as bog-standard Class 10 or Class 6 cards, when used with a good Linux filesystem like ext4. The synthetic measurements of random access or sequential performance don't really capture filesystem performance accurately, although there is a fairly good (but not infallible) correlation between random write performance and filesystem write performance.
The SanDisk Ultra and blue Class 4 cards are also rather good, provided you get one that was made sometime in the past few months – ie. buy one now rather than digging through your spares. They might be available for much less money than the Extremes, especially in higher capacities, but offer nearly as good performance in practice.
There doesn't seem to be any particular problem with Class 10 cards in general, but some specific manufacturers tend to use very low-quality components in the name of minimising costs. The principal offender here is Kingston, but I suspect that some other brands are rebadges of Kingston cards.
Out of the 15 name-brand cards I have tested (in a TrimSlice, which is as close to a real R-Pi as I can get so far), about half of which are Class 10, only a Kingston card failed to complete a simple "format, untar, copy" benchmark run. All fifteen cards (including the Kingston) have repeatedly completed a synthetic benchmark.
Transcend, SanDisk and Verbatim are all known-good brands in my book, but SanDisk is clearly the performance leader.
The key to knowledge is not to rely on people to teach you it.