pieppiep wrote:I think you should always try to not get compressed belts.
If a belt becomes compressed, the production it leads to isn't big enough or you need more belt capacity.
If your belts are compressed or not does not matter at all if you only care about results (i.e. goods manufactured).
The only thing that matters is if your assemblers have 100% uptime or not (for the products that you're interested in).
If you have 10x Assembler-3 making Science-Pack-3, you want these assemblers to run continuously (assuming that you're actually researching something that requires Science 3, of course). If they don't you either have no demand for Science 3, or your assemblers don't have enough materials to work with. That's the end of the story, the belt compression does not matter.
Now to get back to your original question: are compressed belts really good? Of course they are, unless you want to build more belts than necessary. If you need to carry 2400 units/minute from point A to point B, the most efficient way is a compressed blue belt. So you compress a blue belt first before you start to run more belts in parallel.
Also, a belt does not "become compressed", it either is already compressed
at the source or it never has been (the "source" in this case is the splitter that creates the 100% compression, as there is no other way to do that before we get the new inserter in 0.13). What you mean when you say "it becomes compressed" is actually a belt
clogging up due to lack of demand (as you say yourself).