tikz-knots: unexpected(?) behaviour with a circle

Question

Here's my knot:

I started by drawing the red+blue strand, which went quite well with a Hobby curve.

\usepackage{tikz}
\usetikzlibrary{knots, hobby, shapes.geometric}
\begin{document}
\tikzset{
  knot diagram/every strand/.append style={
    ultra thick,
    red
  }}
\begin{tikzpicture}[use Hobby shortcut]
 \begin{knot}[scale=2,
   consider self intersections=true,%ignore endpoint
   % intersections=false,
   end tolerance=0.05mm,
  draft mode=crossings,
   flip crossing/.list={1,3},
   clip width=8,
   clip radius=8pt,
   only when rendering/.style={
   }
   ]
 \strand ([closed]2,0) .. (3, 0) .. (4, 2) .. (1.5, 0) .. (1.8, -0.4)
 .. (2.2, -0.4) .. (2.4, 0.4) .. (0, 2) .. (1, 0);
\end{knot}
\end{tikzpicture}
\end{document}

Then I added in the green link as a circle:

\usepackage{tikz}
\usetikzlibrary{knots, hobby, shapes.geometric}
\begin{document}
\tikzset{
  knot diagram/every strand/.append style={
    ultra thick,
    red
  }}
\begin{tikzpicture}[use Hobby shortcut]
 \begin{knot}[scale=2,
   consider self intersections=true,%ignore endpoint
   % intersections=false,
   end tolerance=0.05mm,
  draft mode=crossings,
   flip crossing/.list={1,3},
   clip width=8,
   clip radius=8pt,
   only when rendering/.style={
   }
   ]
 \strand ([closed]2,0) .. (3, 0) .. (4, 2) .. (1.5, 0) .. (1.8, -0.4)
 .. (2.2, -0.4) .. (2.4, 0.4) .. (0, 2) .. (1, 0);
\strand[green] (2, -0.5) circle[radius=2cm];
 \end{knot}
\end{tikzpicture}
\end{document}

Now, I see:

The crossing numbers have changed of course, but that's not a problem -- I can flip them. The problem is the crossing marked as '2': it's in the wrong place. If I try and flip it, the situation gets worse (zoom from image):

I eventually succeeded in drawing this knot by using a second Hobby curve for the green strand(*). I have drawn my knot -- I'm not looking for further advice on how to do this particular example. But I'm keen to get a better grasp on the knots library and I'd like to understand why the attempt using a circle behaves that way -- does anyone know? I tried making my circle bigger and smaller and shifting it a bit left and right, all to no avail... I don't know if there's something in the knots options that I was missing? Other than where the start point of a circle might lie, it's essentially symmetric with the crossing marked 7, which is handled correctly... so perhaps there's something to do with the circle starts?

(*) The working code, for completeness:

\documentclass{article}
\usepackage{tikz}
\usetikzlibrary{knots, hobby, shapes.geometric}
\begin{document}
\tikzset{
  knot diagram/every strand/.append style={
    ultra thick,
    red
  }}
\begin{tikzpicture}[use Hobby shortcut]
 \begin{knot}[scale=2,
   consider self intersections=true,%ignore endpoint
   % intersections=false,
   end tolerance=0.05mm,
  draft mode=crossings,
   flip crossing/.list={1,4, 5,6 },
   clip width=8,
   clip radius=8pt,
   only when rendering/.style={
   }
   ]
 \strand ([closed]2,0) .. (3, 0) .. (4, 2) .. (1.5, 0) .. (1.8, -0.4)
 .. (2.2, -0.4) .. (2.4, 0.4) .. (0, 2) .. (1, 0);
\strand[green] ([closed]2, -1) .. (3.5, 0) .. (2, 1) .. (0.5, 0);
\end{knot}
\end{tikzpicture}
\end{document}

Answer 1

This is not an answer, but is an explanation of the problem.

When consider self intersections is set to true then each path is split into segments so that the intersection routine can be used to find its self intersections. When doing so, it removes empty segments which consist of just movetos. It would appear that there is a bug when doing this with a circle, since a circle starts with a moveto from the centre to a point on the circumference. By removing the various clips, it is possible to see the path that the knots library is using to find the self intersections:

So when the moveto is removed, it's the wrong moveto. This is a bug. I've opened an issue on github: https://github.com/loopspace/spath3/issues/29