In reply to issue:
The PACI article strikes me as highly speculative.
Assertions of "cyclic loading leads to ..." come from what?
--testing, or musing and finding a way to manually work out a failure?!
I should think that using two 'biners (nevermind the
problematic locking part --fine, if you have it),
opposite-&-opposed, and tying the rope into them
with a multi-(short)-eye Portuguese bowline will do
perfectly well. The multiple eyes are for stuffing the
inside-'biner cavity with 4-5 strands of rope to keep
the 'biners from floating around. The twin 'biners,
even if in gate-loading orientation, should be amply
strong.
If one is sure of the locking mechanism, there are ways
to *snell* a rope to a 'biner's strong axis so to preserve
proper loading.
An alternative approach is to invoke an easy-tying method
(for the adult/trained supervisor) in which a long eye of
<name_your_preferred_eyeknot> reeved through the belay loop,
then through its own eye legs,
turned around those legs and tucked within this turn (a half-hitch sort of thing),
and this eye tip then clipped along with the (twin) eye through the belay loop
--the clipping effecting a simple, low-load locking of the tie-in;
the 'biner run through three parts (eye tip & tie-in twin eye parts)
won't be able to rotate, et cetera (kind of like a cotter pin, in effect).
Snugged down, the structure should hold w/o 'biner,
but the 'biner's there to be sure! (non-locking, fine)
*kN*