Re-formatting pages is just a waste of time (imho). Tomorrow someone can dislike something else... If there will be no overlaps we can proceed with any formatting. There are enough editors to fix articles (at least now). And I'd rather revert any inconsistent changes in markup in manuals without having at least a little hope of fixing the whole manual by the particular editor.
The consistent look and feel of manuals is better than inconsistent usage of the best markup ever discovered (imho).
I'm not sure what you're saying. Yes most people don't follow rules, but it can only help to develop the right rules. The more they make sense and are consistent with what people are used to (italics for variables, kbd as the tag for the keyboard input, etc.), the easier they are to follow.
I can live with bold and italic as your variable text rule if changing it will introduce inconsistency within your wiki. But extending the box and using <kbd> make sense, and over time the wiki will improve.
Anyone got a month or two to spare? Seriously though - since we are all agreed that <code> markup should encompass whole commands, the logical first step is to correct such things as we find them. Even if we felt it worth the time involved, there is no way that you could search for and find such broken instances, so it's down to alertness and common sense.
To be honest, I would rather not include new tags such as <kbd> unless we can see a genuine advantage to it.
One more point. Generally we use wiki-markup (with a few additions thrown into the mix) but as a rule we discourage use of HTML markup as it tends to interfere with wiki-markup in unpredictable ways. For this reason I would prefer not to introduce more HTML markup in the guidelines. Better a new template, if it is really needed.