Since this looks like two ways to do the same thing, while I'm not a fan of multiple images in examples think this is a great use case for it since one thumbnail covers both methods. I'm thinking linkable subsections for:

1. Recommended method

2. Other method where we can write the caveats for when to use this method.

I'm hoping that subsection headings will prime folks to more easily find the one line of code that does the thing that way in each section.

That doesn’t really warrant a whole paragraph and the new axes titles cover this I think.

I think a sentence is fine for a new subsection, doesn't need to be a paragraph. My reasoning for splitting it into two sections is that:
1).the code kinda gets lost/harder to follow the more there is of it. Like here you've gotta match the axes to the method and then find the line so it's a two hops instead of 1
2) provides a way to more clearly identify a preferred/recommended method. I feel like putting em both in the same figure denotes that they're at the same level.

Which like clearly I've got stronger than I thought feelings about this & like I can put in a follow up advocating for this. Not gonna block but also not approving b/c while I agree it's better in that it documents something not documented, I'm worried that adding the second version in this manner makes it a little harder to figure out the "what should I do" from this example.

2. provides a way to more clearly identify a preferred/recommended method. I feel like putting em both in the same figure denotes that they're at the same level.

When I opened the issue I was obviously thinking set_inverted is the "right" way to do this but actually if you want fixed limits then using set_xlim would be the right way. So I'm not sure I'd say one should be preferred in general, but that it depends on your use-case.

I still believe one figure is better, because you have the two code variants right next to each other, which my ales it easier to compare.

we can however make the “two variants” situation more visible in the description.

Will update later.

I still believe one figure is better, because you have the two code variants right next to each other, which my ales it easier to compare.

But here the image is identical, so I'm not sure that helps. And there's a balance in that while visually it may be easier to compare, I think the code can be harder to scan for which piece of code does the exact thing.

So I'm not sure I'd say one should be preferred in general, but that it depends on your use-case.

Sure, but like I think if we're not communicating when to use what, then it leaves the user super confused on why we have two ways to do the same thing.

But here the image is identical

They are not identical, as the fixed limit version has narrower limits. Perhaps that would be more obvious if they were one above each other instead of side-by-side?

Let's discuss based on this version.

But here the image is identical

They are not identical, as the fixed limit version has narrower limits. Perhaps that would be more obvious if they were one above each other instead of side-by-side?

Um these look pretty much identical to me ( realizing cause I had first thought that y was being inverted, even with the title 🙁, so yeah maybe stacked would help - but even then I think it's a bit of a you have to know what you're looking for problem, ie you have to know that auto scaling has margins to really see it in the image.)

I'm less hung up on them being in one figure now cause yeah code that's short enough to fit on one phone screen is probably easily scannable, but I'm still not convinced that there's a benefit either...

Like I think if you're going w/ one figure, then you may as well invert x using one method and y using the other and label accordingly on each axis.

Sort of the point here is that both methods yield similar results: Both have inverted x-axis. The difference fixed vs. autoscaled axes is not immediately visible. We could use a larger range for the fixed-limits case, so that one clearly sees these are not autoscaled, but do we want to draw attention to this?

On one or two plots:
With two separate plots, the code blocks and figures are further apart and thus harder to compare. IMHO this is a disadvantage no matter whether we want to emphasize both variants result in similar plots, or whether we want to emphasize the difference in limits from auto-scaling vs. fixed limits.

Used x, y as variables and "decreasing x ⟶" as xlabel to draw more attention to the x axis.

To put even more emphasis, we could color xticks/xlabel e.g. orange, but I'm hesitant that the additional commands add too much clutter.

Edit: It would also help to be able to highlight important code lines. -> Create a feature request at sphinx-gallery/sphinx-gallery#1283.

the code blocks and figures are further apart and thus harder to compare.

Hmm a point I think I maybe am making badly is that I don't think the goal should be to set up a comparison of results - between the similarity of image and similarity of code, like the visible difference is super nuanced.

The reason I'm advocating instead for the separate sections is precisely to emphasize the differentiation/comparison on use case:

1. invert axes with fixed limits
2. invert axes with auto limits

The above is pretty clear.

If I were doing this,

a) I'd have a third subplot without the inverted axes
b) It sometimes happen that you want to invert the x axis, but it often happens that you want to invert the y axis, at least in oceanography.

@story645 @jklymak I yield here. My motivation and scope was a reasonable quick fix for #28050, which I believe is fulfilled. Feel free to bikeshed further without me.

Let's get this in...