Makes perfect sense. And perhaps I grasped more than I thought I did. I just didn't know I grasped more. This would lend itself to the idea that a CJP must be accomplished on order to eliminate the angular distortion, wherein a PJP (for example increasing pen to reduce the fillet size) would only shift it, though perhaps minimizing to a minimal extent.

With regard to the shortening of the work after cooling I think this will vary to the point of being non existant as the width of the parts increases. In a suficiently wide part the shortening will only ocur in the area that was molten or nearly so. In the case of Kix's "T" joint the distortion occured in the only un-restrained leg of the "T". If an un-restrained full penetration "T" joint was done from 1 side the horizonal member would be warped from cooling stress much like if You ran a really heavy bead down the center of a plate. The non parallel section of the cooling material will pull the verticle member toward the side it was welded from.  

One of my shortcomings is that I have not taken the time to learn how to attach sketches or pictures to my posts.  (If I did that, it might make all this way too easy for me.) Another problem is that I have been very busy lately so I don't have as much time as I'd like for the forum (so if I don't get back to anyone; it's nothing personal).

What I was trying to convey is:  When thinking about how distortion happens, it helps to picture a cross section of the weld and the weld joint, which is also how welding symbols are shown.  The cross section of a joint resembles mitered flat bars- not a bar bent the "easy way" but more like we are trying to make a 2' framing square from 2 flat bars cut at 45 degrees. 

Now, instead of making our square from 2 pieces, we are going to saw a vee cut in one piece that is almost but not quite through; so that we have a "hinge" and can easily push ends where we want, to form the 90 degree angle.  That represents how a weld bends the base metal.
A wider vee has more bend or angular distortion, a narrower vee allows has less angular distortion(AD).

Increasing the weld penetration without increasing the width of the weld face will equate to making a narrower vee in the flat bar; less angular distortion.  Making the face wider without increasing the penetration depth will make the distortion worse.

Making a weld on each side of the joint is similar to making a vee cut on both sides; they tend to cancel each other out to a some extent.  (I'm speaking about angular distortion here, not transverse shrinkage or longitudinal distortion which are similar to AD but are another matter- however visualizing a cross section helps in understanding those conditions as well.)

What is really happening is that more weld is on one side of the cross sectional neutral axis than on the other.  Anything we can do to balance the welds about the neutral axis will decrease angular distortion.   We can 1) restrain the parts from moving,  2) balance the heat input about the neutral axis, 3) increase preheat to reduce temperature differential , 4) reduce the amount of welding and/or reduce the number of weld passes, and 5) preset parts to allow distortion to move the parts to the desired location.   As an example, changing a B-U4 joint to a B-U3 covers numbers 2 and 4.

In the case of a multi-pass weld we also should consider that each pass adds thickness that resists some of the bending from the next pass.  And we need to remember backgouging; not so much the heat input from backgouging but the additional cross sectional area that needs to be filled with weld metal.  An example of how we work with this: for thicker double groove joints (1/2"+), we bevel our material 1/3T and 2/3T groove depths and weld the 2/3T side first.  The 2/3T: 1/3T ratios reduce for thicker and thicker joints but the idea is to have the first pass centered about the neutral axis so it does not add to AD and gives some bending resistance to the next pass.   After backgouging, the grooves for each side end up being somewhat equal in cross section and we have found this method helps reduce angular distortion. 

At the risk of boring everyone to death, another example was some angles we had to make up from 1/2" X 14" and 1" X 5" plates.  The joint was TC-U4b in the 1/2" plates with backgouging required.  There were a few hundred of these to make, mostly of lengths too short to run through our drill-line for hole making.  The solution for short lengths was to make longer lengths that could be drilled and then cut to length, but we knew longer lengths would distort more.  We calculated (fine tuned by trial and error) the longitudinal and angular we could expect and burned our plates with a curve in the length, to compensate for longitudinal distortion.  We did not have sub-arc equpment suitable for the application so we used FCAW mounted on Bug-O machines to help keep the heat input consistent and repeatable.  We used a 40 degree, 2/3T bevel and 0 root gap on the inside of the corner and built a jig to preset the plates at ~95 degrees and a 1G position.  Inside welding ended up being 2 passes (we tried 1 pass but it was not as consistent) and the outside corner was backgouged and illed with 1 pass.  After welding almost all pieces ended up straight and at 90 degrees.  When we started having some that needed heating to correct (close to tolerance but not quite) we also realized the heat number the parts were cut from had a higher yield strength.  An adjustment to the curve fixed that problem.

As much as I've gone on about all this, I think everyone might get a better explanation from the Lincoln Electric publications "Arc Welding Procedures Handbook" (or somethin like that) and "Design of Welded Structures" by Omer Blodgett.  That's where I go for information on this subject.  The calcualtions I use come from those publications.  Also check out the www.jflf.org website, it's got a huge amount of infrmation on it.  (No, I don't work for Lincoln but I am impressed with the valuable information they make available for little to no cost)

(If I've confused the subject for you just tell me to shut up.)