The first “flat” 8 starts to curl / bend, the very moment you make the working end of the second link of the bend ( or of the bight component structure of the eye-knot ) penetrate the two nipping turns of the first. Yes, it is round, and the most important thing it s that is very compact and dense - itsvolume / bulk is very small, considering the total rope length it consumes.
Tweedledoubledee, or Tweedoubledee, for short ! 
This part is actually quite simple when I am not trying to make a video of the procedure. The working end simply follows parallel with the part under the riding turns.
Making the video was rather awkward, so things didn’t flow as smoothly as they normally do. I was sitting on two stools with a tripod between my legs, and the top of the tripod and the camera/phone directly in front of my chest, between my face and my hands. My attention was split between watching the screen and tying the knot. Tying withe the same method and using the same line I normally tie the knot in between 15 - 20 seconds.
When tying with this method, I also find that I can better observe the formation of the second “8” than I can when I tie with the “8” laid out in my hand – the two parallel legs keep me from seeing what is going on, much like the two legs of the collar on a bowline obscure the nip when viewed from the “front”. 
– JP
However, in this case, following my dumb method, one can be helped by the perfect symmetry, and for-see / anticipate what happens to one part of the one link he can not see because it is hidden, by what happens to the corresponding symmetric part of the other link that is shown ( by definition, two symmetric parts can not both be hidden in a “front” and a “rear” view, at the same time ! )
tweedledee bend , and its 99% similar twin tweedledum bend , both easily tied by retucking a what knot 1406.
Nooope, two topologically different knots are NOT similar ! They might even not be similar if they are topologically identical, but geometrically / structurally quite different, like the “bistable knots” (1)
The fact that they can be tied by re-tucking the same knot, is of no importance whatsoever, regarding their “similarity”. See how many completely dissimilar bends can be tied by re-tucking the Reef knot (2)(3), the Thief knot (4), or a certain symmetric Carrick mat (5).
Oh, no need to ignore their similarity 
they sure are not same, but there’s remarkable similariry,
and so they’re named tweedledee and tweedledum,
which means twins , hard to distinguish, by looking.
the diff point is inside the knot knub, an additional crossing
Their claimed “similarity” is only skin deep. I will never call “similar” two knots of different topology ! Is the 0 similar to 1 ? Is the sphere “similar” to a torus / donut ? Especially in this case, the one bend can be turned into a PET eye-knot, while the other can not. Big, HUGE difference ! The “remarquable similarity” and the names chosen by R.E.Miles were unfortunate mistakes ! The Teweedledum ( M. B 24 ) is more “similar” to the "Flat square knot ( M. B 29 ).
The road to call “similar” all the knots that it is hard to distinguish by looking ( but very easy by tying or untying…), is a very slippery one ! Imagine how many completely different knots, indeed, things, you will be forced to call “similar” !
For a FINE, in depth exposition on the issue of the so-called “similarity” of various bends, see :
http://charles.hamel.free.fr/knots-and-cordages/Investigation_6.html
Tweedledum bend is formed by retucking a what knot 1406(grief?)
B29 or abok 803 is formed by retucking a reef knot, very very very difficult to untie
The “Oyster bend”, which is just the Tweedledee bend with the “diagonal elements” on the outer mantle of the knot ( and can be tied with a corresponding tying method ) is much more difficult to untie ! In fact, it is the most tight / jamming knot I know ! So, you see, a “minor” detail can transform a prince into a frog, and vice versa. “Similarity” of knots is a notion that does more harm than good.