I got this right by guessing how the author of the question had misinterpreted the documentation, since the selection of answers made me think that the documentation had been misinterpreted as saying something that it clearly doesn't say. So I think that it's not a good question. Hoever it will probably get people to read teh documentation on this new feature, so it's not a bad question either - just somewhere in between.
The only recomendations in the referenced page are for ranges, not for sizes. Picking the limits (4 times and 8 times) of one of the ranges as two value as best practise instead of recognising that a range of values is suggested is just plain nonsense. Picking the limits of a range which "may be optimal" instead of a different value within the recommended range (up to 16 times) which may be outside the "may be optimal" range is equally nonsense.
All you can say from the documentation about practise is that although you can go up to 32 times it is almost always best practise to stick at or below 16 times - unless of course it turns out that your workload workes better with 24 times. You can also deduce that for many workloads someting between 4 times and 8 times will be best, but that doesn't mean that best practise is to pick 4 times or 8 times, since the clearly stated recommendation says up to 16 times, and anyway picking 4 and 8 as the two best practise values excludes (for example) 6, which is clearly in the range which "may be optimal".