It's hard to get your head around. In essence the board creates 2 curves along its length, sidecut and flex. With the board at rest the SCR axis is in the horizontal plane, and what will be the flex curve is in the vertical axis. Both decrease in radius as the board flexes and rotates along the line of the edge as the turn develops. As the board rotates higher on edge the turning effect produced by the sidecut reduces and by the flex increases. The contributions are roughly even at 45 degrees board edge angle, and flex predominates at higher angles. Remember the turn itself is effectively produced only in a horizontal plane.

A factor is the degree of surface penetration and whether the board can bend enough that the steel edge will engage the bottom of the groove. On prepared race courses and other really firm surfaces penetration will be minimal and the steel edge will bend in a curve with a radius that is very close to the cos(board angle) X sidecut radius. Given that higher board/ski angles lead to more and more board bend and smaller and smaller turn radii, there must come a point where because the ski/board won't bend sufficiently, and while the ends of the effective edge are in contact with the snow the middle of the sidecut lifts off the bottom of the groove. 
At this point on ice the racer might well lose grip.
On softer snow with a deeper groove, the rider merely rides the sidewall with the base of their ski/board and the board flex curve is defining the turn.

I've attached a link to a 2023 analysis of ski turn geometry/physics I found recently.
The author also notes the issues with the cosine function approximation at high angles.
Amongst the titbits in the paper is a comment that on snow testing with the Norwegian National Ski Team found reasonable agreement with the Cosine function sidecut approximation up to about 70 degrees ski angle. I'm assuming they were working on firm prepared training courses.

I'm going to try to use some 3D modelling software to make a visual representation of the curves evolving and the surface penetration. (going to take a while, I'm learning how to use the Blender software from scratch)

Been watching carving snowboard videos where the point of view allows me to see the track carved and/or the base of the board for much of the turn.

There's a point in high board angle turns, even on hard PGS race courses, where suddenly a spray of snow is ejected laterally from the turn groove.

To my eye, in purely carved turns in relatively well compacted snow, it happens only after the board angle is higher than approx. 45 degrees. (In soft snow it starts to happen significantly earlier.)
In racing turns there is often a tail skid or stivot to rapidly turn downhill after transition with a spray of snow, then the edge is set, and then a second spray of snow as the board turns high on edge.

My interpretation of what I am seeing is that the effective G(ravity) force vector for the rider has a more horizontal than vertical direction at this point, and the compression/carving of the snow by the board is now displacing it laterally. If you look at the groove just behind the rider (best seen in the POV in the first video) the track widens with the tip of the effective edge creating the inside curve, the middle and tail of the board riding the outside of the curve, and the floor of the groove sloping gently down from inside to out, while the lateral wall is steep.

That snow spray is probably as good a marker as any of roughly where the transition from sidecut driven turn -> flex curve driven turn occurs.

Plenty of great examples out there, these amongst them.

Stickcam looking back towards rider, from 0:25 has some reasonable track views (better light & grooming)

Tailing cameraman, good views of base and snowspray transition on heelside turns from 0:38 into video (steeper slope)

Any of the runs here from Peking 2022 PGS Finals will show the stivot spray and then turn spray on a firm course.
https://www.youtube.com/live/0yapsVkrJDE?feature=shared

Like the wake coming off a boat hull?

That sidecut radius calculator is great...

https://natewaddoups.github.io/SidecutCalculator.html

• @board-doctor Not really like a boat's wake. That changes depending both on whether the hull is a displacement or planing design, and the speed.
• I was thinking more in terms of the direction of the force vector from the riders COM to their base of support pushing against the snow, and the transition to that pushing more sideways rather than downwards.

@board-doctor 

https://csia.snowpro.com/ServicesServlet/telechargement/document/Performance_Model.pdf

Look at pages 16-17 for a brief discussion and pictures that illustrate the interaction of sidecut and flex in producing the steering angle of a ski (or in our case, snowboard).

A great update from Lars: