It works all the time for me and I've never had a problem with it which is why its the method I advise to use. Show me an example where its broken and I may change my mind or revise the method.
-- which is why I use the pad the bottom of the content wrapper and negative top margin on the footer approach, which is simpler and doesn't break in anything I've tested in.
Yes that's fine, but in practice I find that my clients inevitably break the layout with this approach as they are always adding to the page and add elements under the #content and thus breaking the footer. This has happened a number of times now. With the header approach this has never happened (yet) which is why I prefer it.
Also this approach suffers from the fact that the content area is now 192px larger than it needs to be and any background images or colours extend below the content which is undesirable in most cases.
If you wanted an element in the content section with another background then you would still get 192px of the original background beneath it. Which at that point you find users moving the element out of the content and thus breaking the footer again.
Another point to consider is that if #content has a background and also has position:relative added then once again the footer disappears under the content. Not a hard issue to solve of course but this doesn't happen with my header version and is therefore unlikely to break once handed to a client.
The height:1% on #content is a dangerous approach as far as IE6 is concerned and should there be floats in #content then the user will apply overflow:hidden to #content thus breaking ie6 as the content will be hidden.
The 9999em negative margin is past the limits of old opera versions (early 10 and before if I remember correctly) and unlikely to work and probably why you said you couldn''t get it work before. -32767px is the safe limit and Simon (zcoprpan) from opera explained where that limit came from but newer versions of opera seem to have increased that limit now.
Lastly the position:relative on the html element causes an old safari bug and although its not an issue these days I still think its a hack that should be avoided (older safari would hide any other elements on the page that had position:relative defined).
Taking all that into account I prefer my version but agree that for modern browsers the issues mentioned aren't important apart from the background colour issue and the fact that clients keep breaking the layout when this approach is used.
I would however agree that for pixel precision in a sticky footer you need to use px (although IE6 and 7 will likely always be a pixel out on odd pixel heights)
Your empty content version of the Opera fix does NOTHING here, and not using EM's for the footer (which sucks, but whaddayagonnado) dodges the "Firefox is a retard about EM's" bullet.
It works fine for me in every version of Opera I've tested since we started using it and not seen a problem. I don't doubt that there are bugs in certain circumstances but that's probably true for most everything else also. However, adding a space or come content could make it more solid as Firefox does have a weird collapsing margin bug where floats are cleared using "" without content although I haven't run in to it when used with the sticky footer.