Do you realize that it's $0.09 ? 

You're correct, I forgot that they're grouped inside of file. If seeking is supported then splitting them to different files is unnecessary, I assumed that seek is not supported.

Create a plot [256 kb] -> store plot[0] to plots0.dat, store plot[1] to plots1.dat etc for each 4096 scoops.
Create a plot [256 kb] -> append plot[0] to plots0.dat, append plot[1] to plots1.dat etc
Create a plot [256 kb] -> append plot[0] to plots0.dat, append plot[1] to plots1.dat etc etc

This way there will be 4096 large files (plot0.dat ... plot4095.dat) which can be uploaded to the cloud. You decide how large these files will be, of course they can be split if necessary. They also will be badly fragmented when stored on HDD, but this doesn't matter because they will be uploaded one by one to the cloud anyway.

When mining block 1323, simply fetch scoops from plot1323.dat and so on. 4096 times less bandwidth because unneeded scoops won't be downloaded.

p.s. In normal mining mode only ONE scoop from 4096 scoops in each plot is used. It's inefficient but it was done this way because reading speed is not the bottleneck when using local HDDs and it allows avoiding fragmentation when plotting (leading to faster plotting as there's no defrag step).

As I recall, by tweaking the plotter it's possible to store data for one block per file, creating a set of 4096 files and only reading one of them on each block. So, the required bandwidth can be 4096 times smaller.

Why do you think so? Did he post any code?

In fact I like bitladen's idea of a coin based on CryptoNote+PoC+floating rewards, but imho any of these are too radical changes for burst. It would be interesting as a new coin.

Check the last posts of burstcoin dev, he did not like the idea of changing the block reward: https://bitcointalk.org/index.php?action=profile;u=364787;sa=showPosts

I still haven't got an answer, why do you guys think that tinkering with the block reward would bring more miners? The price will drop even more. The trust will be lost as well because the basic parameters are so easily changed. Developing coin features and keeping the development steady matters more, imho.

This sounds far more reasonable to me:
https://bitcointalk.org/index.php?topic=731923.msg12385695#msg12385695
Maybe the author of that post could lead the development?

I fail to see how increasing the reward will bring in more miners. Wouldn't it only worsen the inflation? Is this the plan to kill the coin faster so another fork can be introduced?

Rewriting the core code to get rid of "NXT fork" fame, probably changing the algo to make plotting much much slower so ppl with petabytes of storage won't be able to takeover instantly, releasing user-friendly mining / wallet software right from the start has a better chance of success imho.

Try downloading and installing JDK, not JRE (it did not work for me, I haven't checked why).
Then open Terminal, cd to wallet folder.
Run
./compile.sh
./run.sh

Btw, on dev2 pool make sure miner sends all shares below target deadline, not only the best one.

Try dev2 pool.

Why not holding the mined coins until they cost much much higher? Power consumption is minimal, hdds lifespan (if configured properly) shouldn't shorten much.

USB2 - ~40 mb/s, too slow for large HDDs. Works well for ~1-1.5Tb
USB3 hdd - 160 mb/s and more

To extend HDD lifespan it's important to disable sleep. If it spins down every block it will die fast. This can be done programmatically for some models or using a hack, for example by writing a file to HDD periodically.

A simple USB fan works well for cooling down external HDDs btw.

Pardon me if it's a stupid question, but how to use this formula?

For example, I have 20 tb plots. What is the chance to find a block solo?

Some of third party tools are incompatible or have known issues, but dev can't be held responsible for other developers code. Did you try dev2 pool? I've checked the shares list and it did register all the shares which were sent to it.