1/About diffraction:
the radius of the image of a ray coming through the lens is : r=1.22*L*O where L is the color wave length,O the aperture.
If you do not want diffraction visible,you have to have r<the width photosite (because of the lowpass filter,we can choose that the radius is small enough instead of the diameter:the diffraction spot will be as large as 2 photosites)
In your case,the CANON 400 D has a photosite size of 5.7 micrometer,so for the red(L= 0.8 micrometers),you have:O<5.7/(1.22*0.

=5.8 and for the blue(L=0.4 micrometers),you have:O<5.7/(1.22*0.4)=11.7
So,in order to be on the safe side,you shouldn't go beyond 5.6,diffraction wise
2/I concur totally with the advice telling you to disable IS while mounted on a tripod.I suppose that you do so in order to get exactly the same picture but with different exposures and it's easier to process
3/If you have something interesting in a near distance,you might consider taking at least 2 pictures with different focusing distances in order to merge them later on with a specific software.In this case,a depth of field compuing software is handy(mine is in my pocket pc).But remember that depth of field depends also on the size you are willing to print and the distance you are willing to look at the print (these 2 last parameters are often fixed in this kind of software,so,I decided I had to generalize it and to add a new feature: if I want everything sharp from here to infinity,how many pictures do I have to take and at which focusing distances)
4/I am in the process of making pseudo hdr landscape(pseudo because I didn't know about hdr a while ago):my way of doing it is to make an horizontal sweep,taking pictures with much ground and no much sky,then,I sweep back,taking pictures with much sky and no much ground.So,exposure is adapted to the sky,for one sweep and to the ground for the other sweep,and,of course,they overlap.To get true hdr landscapes I will have to add more pictures,with exposures really apart from the adapted one what I intend to do...one day.