I invented the ramstage in late 1999 because Linux could not enumerate PCI correctly. Since then, the ramstage has grown in complexity. On one port (RISC-V) I observed that the ramstage was not really needed. In more recent work on LinuxBoot, we've also found that we don't need the ramstage in some cases. If the romstage could load a payload directly, then we'd save some steps.
The romstage on LinuxBIOS V3 could in fact do this, and in this talk I discuss our experiences in restoring rompayloads to coreboot.