The bigger hurdle will likely be the display controller. The TH1520 and JH7110 both use the Verisilicon DC8200, whereas the Spacemit K1 uses a custom display controller that will need its own DRM driver mainlined.

The good news is that firmware is available for that variant. The bad news is that Mesa currently lists the BXE-2-32 as 'unsupported / not under active development' (unlike the BXS-4-64 in the TH1520, which is active). https://docs.mesa3d.org/drivers/powervr.html

So while the kernel driver (drm/imagination) is the right path, the RV2 is a steeper hill to climb: it needs the SoC kernel plumbing and likely some work in Mesa.

Silicon hardware companies have one of the dumbest business models, when it comes to programmable chips. They want to sell chips, because it makes them money through sales. They don't want to spend money on software support, because you cannot link the software to a sale. So software to them costs money and produces no benefits.

But when you think about it even a little bit, you start to wonder. Who is going to write commercial grade software for a commercial hardware product that they don't make money off? Nobody. What you'll get is an anemic volunteer effort at best. The volunteers might even do a good job, but since they are not involved in the hardware development process, they will always be playing catch up and take a year until the latest hardware is supported. So even in the theoretical best case scenario the hardware will be sold when it is least attractive.

This business model is completely illogical and Nvidia doesn't follow it. Instead Nvidia proactively invested in a software ecosystem for their hardware, thereby leading them to their current valuation.

The same applies to Intel and x86. People prefer x86 boxes, because the software ecosystem of drivers, UEFI/booting and so on is fully mature, whereas ARM SBCs are a fragmented mess.