What?  I assume, given your title, that I'm misunderstanding something here...

What you mean when you say the Tarrasch is exactly the same position as the Panov.

Just for reference's sake...

The tarrasch is basically like a reversed panov with black having a tempo down and white has gotten to move Nc3 for free.

The QGA will also usually lead to panov-like positions because black will eventually make the trade c7-c5xd4.

To clarify:

The question about the relationship between tarrasch, panov and QGA is quite a tricky one, however this is how I personally look at it:

In all these panov-like positions time tends to be very important and with just one extra move for black in the main position it turns from a +=/= to a =+/clear advantage for black.

The main reason why cxd4 isn't played all to quickly in the QGA is of course because it will actually directly transpose into a position in the panov where black has played dxc4 to early. Therefore black usually keeps the tension for a while and waits around a bit until white has to make some kind of concession like Bc4-b3 before taking the pawn on d4.

Example on how a transposition might happen:

Playing this way for black is known to give a hard life as white very quickly builds up a dangerous initiative, so, yes, the extra tempo does matter.

So basically my conclusion of the relationship between these three lines is something like this:

Tarrasch defense: Is the same thing as the black side of a panov with a tempo up which allows white larger possibilities to gain a firm control over the isolated pawn and claim an advantage.

The Queen's gambit accepted: Is essentially like a panov where black has played dxc4 to early but with the difference that the c-pawn and e-pawn is still left on the board. This gives black the option of keeping the tension and create the panov-structure when it's favorable for him.

To answer you're second question I think taking up the QGA accepted would be a very reasonable idea as the positions usually leads to ones in panov character. How drawish the positions becomes of course depends on both players, there are lot of fighting lines and drawish lines. However it's generally a quite a bit harder for black to fight for a win if white is all after the draw (notably d4xc5 possibilities) but not impossible.

Unfortunately I'm not the best player to ask about this as I have always found these positions quite unpleasant for the defensive side and I'm even having nightmares of needing to play the white side of the tarrasch.

However during the period when I tried these openings I felt much more comfortable playing the QGA because there are more options to avoid the isolated pawn structure and even when I did end up playing cxd4 my opponents most of the time ended up capturing with a piece on d4 (which is usually not a very challenging approach).

I think the main reason why they do that is because they have actually studied IQP positions (a beginner/amateur would most likely take back exd4 without thinking because it looks the most active) and they knows how one little tempo or slight nuance can bring a lot of suffering in such a position. The main reason why they take with a piece is probably because they are uncertain how that particular IQP position will end up like. There are lots of positions in the QGA where getting an isolated pawn is a bad idea and it ca be rather tricky to know when this is actually the case and the reasons can be very concrete.

As far as I was led to believe IQP positions are strong as long as there are a few minor pieces each, preferably 4; you get the e and c files for rooks also.

One specific difference between Tarrasch and Panov is that in Tarrasch white's most promissing setup is usually considered to be Rubinstein's g3 but in Panov playing g6 means a pawn sacrifice due to white's extra tempo.

I've always found it funny that in the following position 4... dxc4?! appears to be a complete patzer move - after all it surrenders centre and looses a tempo for nothing. Nevertheless it leads to the mainline of the QGA! Does that make QGA unsound?

Well, I've toyed with QGA from time to time in my own games and don't think that it's unsound. I rather believe that 4. e3 by white is such a passive move that black can do better than in QGA. If white plays more agressively black needs to make some sort of concession sooner or late an by making it at move 2 he hopes to have easy development later on. 

I have done quite well with QGA but probably largely due to the fact that I've mostly played it in blitz or against weaker opponents who have not had very good idea how to proceed with white. If white knows what he is doing many of the arising isolated pawn positions require extremely carefull defending by black which isn't easy (see for example Kramnik's white games against Anand). Additionaly, there is of course the "small" matter of 3. e4!?

Obviously 4...dxc4 is not bad because it transposes into the mainline of the QGA. However I would also judge it as a little bit dubious in that exact position as there are better ways for black to punish whites slightly passive e2-e3 move and it seems rather strange to do transpose into the QGA from a repertoire perspective. One of the ways to "punish" white for playing an early e2-e3 would be to transpose into a comfortable version of the tarrasch by immediately playing c7-c5.

It seems to me GM Soltis did a column or two many years ago about the same or similar positions which arise from different openings, sometimes with different theoretical reputations, or with colors reversed, one side a tempo up, etc.  It happens more often than you might suppose.

It might be worth mentioning that already 13... Rc8? in Smyslov-Karpov above is probably a mistake due to 14. d5! and black is in trouble (the idea is 14... exd5 15. Bg5 g6 16. Rxe7+/-)

 Not necessarily, in a position where no side can make progress for example xD