Note that the latest 1. d4 d5 update is relevant to this thread (and to another recent one I recall).

The idea that Black can equalize(/equalise) in the Exchange with ye olde "straight development" (a phrase I recall from R. Fine) incl. ..Re8 and ...Nf8 -- blasphemy!   

Only after reading LeeRoth's first contribution did I realise that the 3...., Be7 move-order effects the choice against the Catalan ! Amazing what you can miss.
However, would the line 1.d4 d5 2.c4 e6 3.Nc3 Be7 4.Nf3 Nf6 5.g3 dc 6.Bg2 Nc6 7.0-0 0-0 nearly bring us back to a 4....., Bb4+ line, not considered very promising for White ?

In answer to that last bit, I'm pretty sure that GM Williams didn't mean to say that 3..Be7 literally "forces" 4.Nf3.  He probably meant to say that 3..Be7 induces or encourages 4.Nf3, which is, after all, the reason behind 3..Be7 and was more or less the conventional wisdom for a time.  These days, it might not be as true as it once was, since 4.cxd5 has become a major option to 4.Nf3.

For years, I played 3..Be7, with the idea of avoiding the Nge2 Exchange.  Today, I switch off between 3..Be7 and 3..Nf6, as I am just as happy or unhappy to face the Exchange after either move.  Theoretically, I don't think one move is any better than the other, so for me its mostly a matter of mood, how I'm feeling, etc.

If you are trying to decide which to play, I would suggest trying both and picking the one that you like better or that makes more sense to you, regardless of how many GMs play it or the evaluation that your engine gives it (within reason, of course).  Better to know what to do in a position, than to be theoretically better/equal and have no idea.          

         

Talking about flexibility, I suspect that 3...Be7 is indeed more accurate if Black is dead-set on playing QGD. On the other hand, 3...Nf6 may offer more options for Black, for example if White plays 4.Nf3 then Black has the extra options of Ragozin 4...Bb4 and Vienna 4...dxc4. 

This begs another question: after 1.d4 d5 2.c4 e6 3.Nc3 Nf6, the opening trees indicates that 4.Nf3 is predominately more popular than either 4.Bg5 or 4.cxd5. Is this really the case when GMs approaches this position, or just because the opening trees account for the transposition from the sequence 1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.Nc3? Interesting to know.

GM Simon Williams, when commenting on Isle of Man 2016 Round 1, suggested that 3...Be7 forces 4.Nf3. I am not sure if it is a slip of the tongue, since 4.Bf4 and 4.cxd5 are certainly possible, but maybe there is some reason for Simon to say that.

By the way, this reminds me of the latest 1. d4 d5 update, which addresses the old-school ...c6/Be7/Nbd7/0-0 Exchange with Nf3.  I wonder if there will be a follow-up about Nge2.  (I've only seen the summary stuff.)

I heard from many sources, like Seirawan commenting in Sinquefield Cup and NM Dereque Kelley in one of his youtube videos, suggesting that after 1.d4 d5 2.c4 e6 3.Nc3, 3...Be7 is more accurate than 3...Nf6.

I understand that the main problem for 3...Nf6 move order is the Exchange Variation 4.cxd5 exd5 5.Bg5, which is precisely the 3...Be7 move order avoids. Although after 1.d4 d5 2.c4 e6 3.Nc3 Be7, White can still go sharp by playing 4.cxd5 exd5 5.Bf4 c6 6.e3 Bf5 7.g4.

My questions are:
1. If theory suggests 3...Be7 is more accurate, why would anyone still plays 3...Nf6? When I was checking the latest database, 3...Nf6 is still more popular than 3...Be7.
2. Why would any GM willing to defend the Black side in the QGD Exchange Variation after 3...Nf6 4.cxd5 exd5 5.Bg5? I know Black is solid, but theory considers Black passive. 
3. After 3...Nf6, if the Exchange Variation 4.cxd5 is so good, why would any GM opting for 4.Bg5 or 4.Nf3?

Many thanks you for your time