Yes, you're right about the builtin bootloader. You must also force a few pins on the device to enter bootloader execution mode; this is what the 74xx125 tri-state buffers shown in Dallas' documentation are for.
I do not know the device, but would it not be possible to make a NoTouch for it?

If that is really the case than this should be the device of choice for students and hobbyists IMHO.
for that I would much more suggest the SILabs offerings (except f3xx) since these have a built-in (free) ICE capability, and I can see no better way to learn/study than to see what goes on inside the chip for each and every step.

If you can run it at moderate frequencies, it's quite manageable, and can still outperform a lot of more conventional 8051-based devices. But, as you head for the upper range of clock speeds, you start running into issues.
This problem is - to some extent - true for any clock speed. A device capable of high speed must - by definition - have steep rise and fall. Steep rise and fall can create havoc on octupus wiring regardless of clock speed.

Erik