What I wonder is: why 4 ports? Maybe 3+2 or whatever.
Part of it is determined by the I/O capabilities of the CPU/GPU - especially mobile chips. The CPU either has to have embedded USB/Thunderbolt controllers or to provide enough external PCIe lanes to drive external controllers. If you want to connect 4 x 4k-or-more displays, then the GPU has to (a) provide 4, 4k-capable DisplayPort streams and (b) have the grunt - and enough VRAM/Unified RAM - to smoothly drive that many 4k+ screens using a graphically rich OS like MacOS. All of that uses space, costs money and (in particular)
uses power and generates heat.
The M1 MacBook Air and iPad Pro deliver amazing performance and battery life and run without a fan - that didn't happen without compromise. Only supporting one external display, "only" two Thunderbolt controllers and just enough USB to run the keyboard, trackpad, webcam etc. is sensible for a system-on-a-chip designed
primarily for fanless ultra-portables and tablets.
I don't think the specifics are known, but it's a reasonable educated guess that the M1 chip only has connections for one external DisplayPort stream (for the laptop display), two TB3/USB4 outputs (sharing a second internal DisplayPort stream) and maybe 2-4 USB ports - which the laptops will use for webcam, keyboard, trackpad etc. (If you have an M1 Mac you could do a system report and see what's connected to what...) Maybe - like Intel - they're in the form of universal I/O lines that can be configured to various permutations of PCIe/USB3/etc. by the system builder, but that's the sort of complexity that it might be sensible to leave out when you're making your own bespoke chips.
The M1 Mini doesn't have webcams, trackpads etc. or an internal display, so it would make sense that it has a couple of USBs and a display output to spare. The M1 iMac would have some spare USB (no keyboard/trackpad) but would need its internal display connection (so no chance of HDMI). As I said, nobody who is telling knows the exact M1 specs but the machines released so far are
consistent with being determined by the I/O resources of the M1 itself.
Looking at the 2016 MBP design, Apple probably
couldn't have added any ports
while keeping 4 TB3 ports, even if they'd wanted to. There probably just weren't enough PCIe/USB3 lanes coming from the chipset (...and they had the T1 chip/touchbar combo to drive, too). What a lot of people would have preferred would have been 2 TB3 ports + the old complement of USBA/HDMI/SD (which, together, would have used less resources than the second TB controller).
...that's the problem with TB3/USB-C: each fully-featured TB controller needs a 4xPCIe connection, two DisplayPort streams and to be hooked up to the power/charging circuitry. That's a
lot of resources that need to be provided by the CPU and a lot of extra cost/complexity/power c.f. a couple of USBA's and an HDMI. If you cut corners on that, then you end up with different supposedly "universal" ports having different capabilities and having to break out the Advanced Dungeons and Dragons handbook to work out where to plug your kit - something that Apple likes to avoid (although not always - see the 2016 4-port 13" MBP where 2 of the TB3 ports had reduced bandwidth or, of course the M1 iMac.)
Of course,
some of it is Apple's combination of form-over-function obsession and penny pinching habits. Now, the
entry-level M1 iMac with just the 2 ports and no ethernet is just pure pricing strategy on Apple's part to create an artificial distinction between the $1300 and $1500 models. There's no practical justification for not putting those ports in the $1300 model. Plus, throwing a couple of the most commonly needed dongles in with the 2016 iMac wouldn't have
solved the problem but it would have calmed a few tempers...