A short script running getblockbycount revealed that the highest combined fee so far was 6.71 BTC in block 55812
http://blockexplorer.com/b/55812
In more recent times, there were a few blocks with fees above 2BTC. This one is the most recent:
http://blockexplorer.com/b/120528

I may be interested. A few questions:
1. When would it start? Do you have the machine or you'll need to buy it and setup first?
2. Are you going to pay daily amounts that can be calculated from difficulty or is it going to be a fraction of your daily pool payouts?

The trust is of course a problem. I would be fairly easy to take the 300 BTC and run. Or even better. Take 300 BTC, pay it back for some time, gain some trust, sell 10 times more contracts and run with 3000 BTC. All without owning mining machines. It would help if you indeed owned the machine and could show some income stream.

You yourself. The very first transaction in the block is the generation (which gives 50 BTC to the miner). The address is one of your own in the wallet that was generated randomly.

Because the address is essentially a very large number, there is an astronomically small probability that two people are doing the same work.

When mining, the server makes very little more work above the standard non-mining bitcoind. My miner used 20 minutes of CPU time for the last 3000 minutes of uptime. I have severals miners with getwork every 2 secs submitting difficulty 1 solutions which the server verifies (as mostly above the target). Most of the work a deamon does is verifying transactions which is done regardless it is doing getwork or not.

When mining solo, be prepared for very long spells of no blocks from time to time. 3-4 times the average is not that uncommon. And it can be quite frustrating. And sometimes you can mine 4 blocks during the time one is mined on average. Poisson distribution has a very large standard deviation.

Depends on how do you define "a dent". The Bitcoin hashrate increased by about 1000 6990 during the last 3 months. It's about 0.1% of the AMD graphic revenue during a quarter which is around $400 million. 

Of course not all of the new mining capacity went to AMD (some were old purchases, or nVidias or ArtForz's ASICs) but all 6xxx purchases and a large part of 5xxx went directly to AMD.

If the difficulty increases 10-fold, which is highly likely, the increased capacity will be 1% of the AMD GPU quarterly sales. It will start to be a non-trivial part of their market.

But the merchant is not required to return the shipping charge.

When Bitcoin matures and volatility drops, a 7-day call option will be much less valuable than the shipping charge.

The difference between 1Gb/s and 1Mb/s will be virtually nil but I'm pretty sure a 14.4 kb modem will sooner or later show it's impact. If you mine, you need to have the current chain. If you are slow on update, you will work on stale blocks for some of the time. Some of the blocks are large, a 100 kB block will need a minute to download and during this time, you are doing nothing useful (but I'm not sure if the correct block header is enough to update the previous block hash and the mining can start on the current block even without full block download so the impact will be much smaller). And if you find a block, you may not be the first to broadcast it to the world. With slow upload, there will be many miners with faster connection who find the block with the same root and broadcast it to the world faster. And then you'll lose the race. It's hard to say without some more rigorous calculations but I wouldn't be surprised if miner with 14.4kb lost 5% of it's hashrate on stales and lost races.

Catalyst drivers are a major pain in the neck. They are frequently completely broken. I've written this guide with Catalyst 10.12 and SDK 2.1 and I recommend against any other versions unless one really has to (e.g. 6990). I used Ubuntu 10.04 but 10.10 worked for many as well.

BCEmporium,

If you want to get rid of this awful deflationary BTC, I can exchange them for testnet coins for you. You will have a part of an alternative blockchain right now. Exchange rate: 500 testcoins for 1 BTC. 10000 testcoins available. You can call it BTC/silver (or bronze).

Why do you think so? This is a transfer from a wallet that contains generation from luke-jr pool that creates amounts with 8 digit precision:
http://blockexplorer.com/address/14rJYeAPkXGDRVeTcEn7czE9xNaU5aRi64

Nothing sinister is going on. And even if so, why should we care? Bitcoin cannot hold much of entropy and even if such trasfers would become more common, it's not a bigger problem than transaction spam.

All the gold ever mined from ancient times is worth about 8 trillion USD at the current market prices.

Facebook (which will probably be uncool and likely much less popular in less than decade) is apparently worth 50 billion.

There are (usually) two outputs to each transaction. One is the change.
https://en.bitcoin.it/wiki/Transactions

If you pay $20 with $100 bill, you are also getting change. This is how it supposed to work. The change you can also spend.

I don't know what was the original Satoshi's plan but 1MB block size limit and 0.01 BTC fee per kB is just 10BTC. There has to be something left for the miners after the block reward is lowered and finally gone. If the fee was 0.001 BTC, the standard reward would be only 1 BTC. It is a minimum fee, so it can be higher if the transactions are very popular and larger than minimal fee is required to be included for the block. But there is a balance between the users and miners and any change on fee rules must be very careful not to break this balance.

Not really. It took 3 months to be ahead (late January to late April is 3 months on my calendar) but poclbm caught up in early February (4th to be pricise), i.e. in 10 days from your first hdminer post.  For 5970, I got 570 MH/s at stock clock with the February version and there were a lot of reports of 565-570 MH/s at stock clock depending on SDK/system/driver/options.  Earlier poclbm versions were a few percent slower so hdminer was an improvement at that time. But only for 10 days.

Assuming Phoenix (or any other open program) will not catch up. For 5xxx series, this assumption has proven false fairly quickly.

20 BTC. I can start right away. I have an idea what is wrong but I cannot promise I can fix it. I can fairly quickly (say half an hour) decide whether I would be able fix it or not. If not, you'd owe me nothing.

I'd need root access so either you'd need to trust me or you would reinstall it from the scratch.

Did you use any guide for the installation?

I'll do it for free if you let me copy your wallet.dat

Seriously, it's hard to say anything if you don't say what these "errors" are.

You essentially  strengthen my argument because even if non-constant-supply stuff can have a bubble, Bitcoin can have it much easier. "They are not making land any more" was the housing bubble main argument.

But there are other examples. There was a bubble in historic coins and stamps which after minting/printing have constant (or falling due to loss) supply (these cards or dolls were also finite supply after producing one series)

Face it. Bubbles are created by human greed and may only be enhanced by cheap credit but the root cause is the human nature itself.

No. Easy credit helps in creating large high-impact bubbles but a bubble can happen with just irrational exuberance. There were several bubbles with zero credit involvement.
Baseball card bubble:
http://www.slate.com/id/2247677/
Beanie babies:
http://seattletimes.nwsource.com/html/businesstechnology/2002020751_beaniebabybubble31.html
and many others.

Bitcoin could be no different.

Sendmany works. One of the pool operators (slush I think) uses it for pool payments. And there is always testnet if you feel uneasy.