3M would be rather awesome. If so the Switch would be sitting at 136M sold and ~138M shipped (Nintendo Q3 shares figure).

2023 has sold 12.55M Switch units (minus december) with ~2.54M sold in Q3 (minus december)
2020 has sold 22.25M Switch units (minus december) with ~5.81M sold in Q3 (minus december) and 12.01M sold in Q3 with december.
If the Switch sales 3M in december, its would sell 5.54M in Q3, 45% of 2020's Q3

Just looking at the VGChartz Charts and it's looking like Xbox One's 2018 Q3-december (2.3M), and 3DS 2012 Q3-december (2.49M).
Xbox One 2018 Q3 ~4.8M (2.49M december), this was Xbox One's 5th holiday & their first party AAA holiday game was Forza Horizon 4
3DS 2012 Q3 6.43M (3.94M december), this was the 3DS's 2nd holiday & their first party AAA holiday game was Paper Mario: Sticker Star

Honestly I'm trying to figure out what was Nintendo's first party AAA holiday game.
Oct- Detective Pikachu: Returns, Super Mario Bros Wonder
Nov- WarioWare: Move It!, Super Mario RPG

I assume its Super Mario Bros Wonder but Oct felt a bit early for a flagship holiday game.

Its incredible that anyone would suggest Nintendo should not launch in 2024. They should have already launched last fall. What type of a tech company waits until they are in significant decline before releasing new hardware?  Apple? No. Sony. No? Samsung. No?  If anything Nintendo should be releasing more often. Its not like the old days when Nintendo had two seperate systems and their handheld were always selling well. Nintendo's revenue is way down in 2023 from hardware sales and that shouldn't be the case. No need for it.

What's interesting is that some of the best years in hardware (and presumably software sales) for the PS1 were later in its life. 

By March 1999, PS1 has shipped 50 million units (slightly under half of its lifetime sales). By October 2001 (a year after the NA PS2 launch and over a year and a half after the Japanese PS2 launch), PS1 had shipped 88.26 million consoles.

PS1 sold more consoles when PS2 was in the serious planning stages and once the PS2 launched than when the PS2 was barely a blip on the radar in Sony's plans.

Now, even if Switch 2 would've launched in 2023 there's little chance by percentages Switch would've pulled off such a feat like PS1. Switch has gotten closer to its saturation point than PS1 had. 

But yes, I don't see why people thought 2023 was too early for Switch 2 when hardware was declining (but still great) and software sales won't die when Switch 2 comes out. Shoot, in the likely chance Switch 2 is backwards compatible software sales will still be high.

2024 isn't too early at all, and 2026 or later is not a good idea at all. 

Since the release of the Game Boy, has Nintendo ever had a year where it moved less than 15m total systems? Even if Switch 2 releases next March, still within the next fiscal year, 10m total units for this calendar year is just not gonna cut it when it doesn’t have to.

Nintendo has had plenty of time to prepare for the launch of Switch 2 and it only makes sense for its new console to arrive this year. Sales this last holiday quarter (particularly in Japan) speak for themselves. And that was with a banger of a year of releases too.

Personally, I’m expecting an announcement sooner rather than later at this point; possibly by March, with the new system releasing between June and September.

And it doesn’t have to be curtains for the original Switch at all, as Nintendo has had plenty of experience supporting two systems, after all. I expect it to be around for a while, albeit with modest numbers, but still making money for the company with games and subscriptions.

A little bit of knowledge is lacking on your behalf in regards to fabrication processes here.

The vast majority of chip manufacturing in the world is actually on older process nodes, there are fabs that are still producing 90nm chips for IC's and Industrial, Audio, Networking (Think Bluetooth and Wifi), Chipsets and more.
In 2022 for example, TSMC actually made 25% of it's profits from 40nm and older nodes.
https://www.anandtech.com/show/17470/tsmc-to-customers-time-to-stop-using-older-nodes-move-to-28nm

Where other companies like UMC, SMIC get 80% or more of their revenue from older nodes.

Global Foundries actually stopped trying to build capacity on latest-generation nodes and has stuck with specialized 180nm/130nm/110nm/90nm/65nm/40nm/14nm etc', because the profit was greater.. And actually worked with Google to expand it's 180nm capacity.

The other misunderstood part of fabrication is that a plant can actually produce chips on different geometries.
For example Global Foundries Fab 7 produces chips at 130nm and 40nm on CMOS and SOI on 300mm wafers.

ON Semiconductor actually increased 65nm and 40nm capacities.

In 2023 India finished dumping 3.2~ Billion dollars on a new fab that produces 45nm and 65nm chips because it's actually a really good node with really good electrical characteristics for analog chips  which is ideal for defense, space, industrial and automotive sectors.
https://www.theregister.com/2022/11/28/indias_first_chip_fab/

So just because it's an older lithography doesn't mean it's useless of wasting resources, older fabs don't disappear the chips they produce just changes or they get retooled to add more capacity/lithography/patterning etc'.

So the old fabs for the Tegra X1 are not going anywhere, once Tegra X1 is done and dusted in the Switch, nVidia will still be producing Tegra X1 for years to come for industrial/specialized purposes within the confines of current contracts who sometimes require support for decades.