If your host machine is a real server (eg poweredge), itll do a selftest. This already takes tens of seconds. If you want fast bootup times, you need to be the BIOS. The example in the top post is stuff that either loads data off a flash chip (like a bios works) or a disk (requires some bootstrapping).
Some of the companies' servers (e.g. Dell) can boot pretty fast in these days. If you want see something that boots slowly, try an IBM HS22 blade (3+ minutes self-initialization after insert, 2.5+ minutes to pass initial BIOS screen to storage initialization and boot-start).
IBM's QPI-linked dual servers boots even slower as one technician explained to me. Presumably you can make coffee during the wait.
What does self-testing cover? RAM? That's something the OS could do as a background task, incrementally making memory available during boot after a chunk has been scanned.
DRAM testing is not really something you can do incrementally. Localized, bit-level failures are only one kind of failure; things get really interesting when you have disturbances (e.g., a shorted address line or something causes writes to some location to also change bits in another location).
Also, ECC failures usually cause a machine check. Not sure if you can control this on modern machines, it might be all or nothing.
