Only because software RAID works amazingly well! And now with SW RAID, when a disk controller goes bad you aren't screwed in cases where the manufacturer no longer makes them.
Does it though? As someone who used MD raid for absolutely everything for over 10 years, were now switching to HW raid. EFI / GPT disks just don't work that well with OS level raid, because the motherboard is well within it's right to modify the EFI partition, which if you have raided at the OS level will cause massive issues.
The EFI partition doesn't have to be RAIDed. All you need in the EFI partition is a stateless, RAID-aware bootloader that is hardcoded to look for a RAIDed volume matching a generic label (such as "rootfs") and then chainloading the actual bootloader from there. Then you can just install this "shim" bootloader into each disk's EFI partition once and you don't need RAID or to even mount it ever again.
Typically random writes must first set a parity in doubt bit. In HW RAID that doesn’t require an additional write to a metadata location, but is instead maintained in some nv memory. Similarly, a read modify write in software raid-5/6 have some poor trade offs. It’s common for a random write to perform a read of both the destination location, as well as the parity. This is followed by a write to the destination blocks and the parity blocks. What happens if power loss happens on the write? Maybe you get a torn page. HWRaid can protect against these surprisingly common failure modes. There are file systems which aren’t impacted by these failure modes, and software raid would be fine, but anything performing random writes, HWRaid has its place.