This loop can be dressed in 4 forms, not only 2. ( See the attached pictures.)
The dressing shown are not uniform, and there are more
than four, by this measure. I.p., the eye bight of the top
image should be rotated 180deg (top leg away, lower leg
coming towards viewer) and resemble the 2nd image
--the difference being which of the collars the right-side
end flows into. (In fact, one can see in "in-the-wild" images
of
fig.8 eye knots (not "inline") dressings that fall out of
symmetry like the top one; this might be induced in part
by torsion in working with the eye bight.
In a tight enough knot, the three encircled and nipped segments
( the two legs of the bight and the "lower" standing part )
will be forced to settle in a "dense" configuration, in a position where
they will touch each other. ( So, they will not remain "in a row",
as shown at the fourth picture above ).
And the turn of the eye bight's (mere) two, twin parts
won't lie in a plane containing the axis of tension
--but the
fig.8 knots are often illustrated in this way.
More to the point, practically, is that one should be able
to position the *tail* (where the eye alone is loaded in
opposition to the intended end) AND THEN set it into
place by tightening the knot (where, were it loose, that
tail could conceivably be pulled out of that position).
NB : It is most interesting to read test data presented
in the
CMC Rope Rescue Manual (3rd ed.) (<c>1998, p.56)
shows
this "inline fig.8" to be weaker than a (normal)
fig.8 when through-loaded !? Granted, through-loading
gets one a sort of bulked-up
square/reef knot, but the offset
loading of a
fig.8 isn't typically regarded as strong
--and it known to be dangerous for security (but by this
data, clearly it was amply secure, and strong!). The
respective figures are : 59% & 65% ; the
butterfly(in a dressing which we don't know --nb!) was 69%.
I think that the second leg of the bight ( the leg which is not a direct
continuation of the "higher" standing part and the fig.8 knot ) is not
constricted hard enough- and this happens in all the 4 forms of this loop.
If the "lower" standing part is not loaded, or if it is not loaded sufficiently
enough, this leg may slip.
Rather, with the same observations, I would suspect
the
eye leg would possibly pull out --consider that
loading it delivers 50% to this part, loading the ends
puts 100% on the connected end, though that is
mitigated by going around the collar; but this loaded
end also might help mitigate the other end's turn
& nip, too.
--dl*
====