Surely there will be companies selling PCI-e dedicated hashing devices to amateurs?

From reading the forums I get the impression it's the other way around. With Windows there are a lot more headaches, difficult to configure and clock, problems starting up cards without a monitor attached, restarting for updates, limits on the number of connections, etc. Linux seems much more sustainable. Also Windows costs money (yes, I'm one of those).

I know I could figure it all out myself with some research, but I know I'll mess everything up and spend hours on trial-and-error. And it's not just for me, I'm sure there are many other people for whom this guide will be very useful.

Let's wait and see who else is up to this . I think it's better to have someone with a lot of experience with this to make the kind of coherent guide I'm looking for.

Collusion is difficult with hundreds of competitors.

I am offering 25 BTC for writing a comprehensive, easy to use guide to mining on Linux, and putting the guide on the wiki. It should allow someone who knows very little about Linux to configure a Linux system (preferably Ubuntu) to serving as a successful dedicated mining rig. It should be compatible with 2 X 5970 setups.

If such a guide already exists, I am offering only 5 BTC for finding it and putting it on the wiki (I searched and didn't find anything, on the wiki or elsewhere). If existing guides fall substantially short of the standards I require here I will offer the full 25 BTC for modifying one to satisfy the requirements.

The guide should cover:
 - Installing a Bitcoin client, and configuring it to receive RPC connections from the local machine and other Linux/Windows computers on the LAN.
 - Installing ATI drivers.
 - Installing the most suitable version of ATI stream SDK (including a recommendation for which version that is, and explanation why).
 - Installing a miner.
 - Disabling Crossfire on the 5970 (if indeed that's a step that should be taken).
 - Over/under-clocking the cards.
 - Running the miner on all GPUs, complete with which flags should be used.
 - Setting the Bitcoin client and the miner to automatically run on startup.
 - Troubleshooting the most common pitfalls (if there's no reason it shouldn't work out of the box, no need to make up problems to satisfy this).
 - Any other crucial steps that I don't know are necessary.
 - A link to a guide about using VNC or else to connect remotely from a Windows computer. (This looks pretty good, other suggestions are welcome.)

When I send a graphics card to a shop for replacement under warranty (I gather that it's common in some places to RMA it directly with the manufacturer, but not here), can they tell if I overclocked it? Will they cause problems if they determine that I did? What about underclocking?

In shared-based, all shares have the same weight (that is, portion of the block reward).
In score-based, recently submitted shares have more weight.

Let's call it "Bitliquid". Since liquid is uncountable, we'll measure it in liters, and talk about sending 21.5 liters of bitliquid. Divisibility problem solved.

Keychain?

It's like lording land. But with a potato.

When you submit a share to a share-based pool, your expected payout from it depends on the number of shares already in the round and is much higher in the beginning of the round. Cheaters who are after a profit but have no interest in otherwise causing harm, will mine until the number of shares is 0.435*difficulty, and from that point mine solo or switch to another pool.
If it helps your intuition, think of this - if there are already 2*difficulty shares, then no matter what happens, any share you submit will give you less than 1/(2*difficulty) of the block. While if you mine solo, you have 1/difficulty chance to get a whole block.
In the limit where there is a nest of cheaters that supply 100% of the mining until the point of improfitability, the honest participants are left to split the rewards of rounds longer than that, which on average is 70% of solo mining payout. This is much worse than any fee, even the seemingly high ones for PPS.
This problem can be mitigated by making it hard to find out how many shares are in the current round. I don't know how effective this is.

Right, sorry if I made it seem like it does. I just wanted to note that your (molecular) list of currently accepted methods was incomplete.

Pay per share is good if you want to be able to very easily verify that you're getting exactly what you're supposed to and not being cheated by the operator or other participants, so it's a good way to start out. But in a few days you'll outgrow it and want to maximize your earnings. At that point I recommend score-based over share-based. Score-based has some more variance, but higher expectation if cheaters are present (or the same expectation if there are no cheaters).

slush is the only one I know to use score-based. Hopefully at some point someone will implement my method, which is not only cheat-proof but also has controllable variance.

It's just not the same when you're consistently progressing towards a goal with a known finish time. Sleepless nights of wondering "did I find a block just now?" for the win!  

I confirm that (but apparently they also accept cash). Also, for convenience of fellow non-German speakers who would also like to take a look, here's the Google Translate link.

See my earlier comment for how I think this should be done. Just for fun, here is what I think would be a very bad way to approach most shops, where actionable information is eschewed in favor of zealous ramblings of details which are irrelevant at best, suspicious at worst. (With some modifications, it could be a good pitch for organizations that already strongly believe in Bitcoin's values).

I think MyBitcoin have currency conversion built in to their merchant services.
A few getting started guides for merchants have been written, for example this. There is some work to be done on making something more comprehensive. There's also this which has a different focus.

Good idea. Though I think I'll sit that one out, seeing as I only learnt about ScummVm and SCUMM yesterday after looking at the DOSBox news. This is despite the fact I have a special place in my heart for a few of what I now know to be SCUMM games, like Loom, MM2:DotT and Hero's Quest.

It's already pretty much pointless. There's no sense in deluding newcomers into thinking that by taking the effort to waste power on this they are supporting the network in a non-negligible way. Humans' incentive structure is weird, if you tell them something will "contribute" they won't bother to ask "how much does it contribute?".
On the other hand, when a person contributes something insignificant (or believes he does) it causes him to be more responsive to making substantial contributions later on. This is an argument in favor of including the "feel good" button, but not strong enough in my opinion.

We shouldn't assume, and it looks like Qbix might be dutch, but they currently solicit only USD donations. It's not guaranteed they can convert BTC directly to their local currency (I know I can't). They will probably be satisfied to know they can receive the more familiar USD. Also, should they ask, I'll recommend them to use CoinCard which only gives USD. Being able to convert directly to their required currency to save some exchange fees will be useful to them later on, but at this point it looks to me like unneeded clutter.

You can have a look at what I am going to send to DOSBox here.
A few points which are important in my view are:
1. Don't be a spammer. If it's a shop, only contact them if you've made purchases before. If an organization, show that you are actually involved in and care about their cause.
2. Unless they are very much inclined towards the philosophy of Bitcoin, they won't care how great it is to have an Open Source peer-to-peer deflationary cryptographic currency which is not controlled by any possibly malicious single point of failure and that they will be helping promote a better future. These can be mentioned to provide some perspective of what Bitcoin is, but downplayed. They care about what's in it for them, how it will help them in their own lives. They want to know that Bitcoin will give them extra business/donations they can convert to fiat currency which is of known use to them. That's not them being selfish, just practical. If they later choose to be more involved that's an added bonus.
3. They will want to know that with Bitcoin, people can send money easily and with no/low transaction fees.
4. Shops may be interested to know there are no chargebacks with Bitcoin.
5. Make it personal. Let them know how you personally would love to pay them with bitcoins.
6. They will be concerned about the work required to receive Bitcoin payments. Let them know that it's easy and offer to help (it helps if yourself you know a thing or two about Bitcoin and its related services).
7. Be polite, of course.
8. Don't overdo it. They will be put off by too much enthusiasm.

Update: 2 organizations I've contacted were concerned about what they could do with bitcoins, so I recommend mentioning CoinCard in the first message (for shops, too). 2 shops were concerned about the work required to integrate Bitcoin to their shopping carts, so you may want to familiarize yourself with existing solutions to make it easier for them.