I’m just defining a term, hence the “or whatever”.

But personally I find Photoshop to be more flexible and more capable than any of these film tools I’ve looked at, none of which seem to be capable of the kind of (very non-standard) workflow I use in Photoshop. Photoshop tools (layers, masks, adjustment layers [especially curves, and more recently full color lookup tables], blend modes) can be combined in very sophisticated ways which can be set up via a macro action, whereas most of these other tools are a bit more prescriptive. For example I can build my own “hue vs. hue” tool (from the Resolve marketing video) out of more basic Photoshop building blocks, and set up an action to pop one up with a keystroke. To be fair Photoshop also has piles of highly constrained single-purpose tools, most of which are outdated and relatively useless.

But none of these tools was ever designed with a user in mind who would think abstractly about building blocks and means of combination. More abstract and generic functional-boxes + arrows kind of interface could probably do better from a flexibility perspective, but can be a pain in the butt to use.

Someone with a color science, user interface design, and art background and a lot of time to experiment who was trying to make a tool for professionals could also certainly do a heck of a lot better. Most of the individual user interface components in these tools [including photoshop] are pretty uncreative and inflexible. Maybe this will be me in a few years when I have some more time, we’ll see.

Admittedly I deal with still images, not video.

Resolve specifically and PS are operating differently. Resolve's color correction is nodal-based, nodes stack by coefficients, not output (like layers generally do). The usual example to show this off is to drown the blacks in one node and pull them back up in the next, without any loss of detail, since Resolve combines the coefficients not the outputs of the transformations.

Resolve exactly lets you put your layers in graphs instead of have them linear. It also lets a later layer bring back detail that was removed in a previous layer, while Photoshop doesn't. Here's an example: https://youtu.be/YxUoW5_gMjQ

That video shows Resolve to be a pretty capable tool. (Can the nodes take custom user-specified logic in them, e.g. some GPU shader or the like?)

But everything shown in that video (other than motion tracking) can be done step for step equivalently in Photoshop, albeit some bits take some nonstandard combinations of tools. Some of it would probably be more convenient in PS to achieve via a different method.

I don’t exactly understand what the compositing order is of “parallel nodes”, but I’m quite convinced I could mimick the effect almost precisely given a technical description.

The nonlinear ordering you mention could be done using smart objects, but that can get annoying, so it’s usually easier in practice to just duplicate the layer (which breaks the link if you change something up the compositing chain later).