I know ULINK2 can do this (a features table of ULINK2 showed), but I don't know what catagories of 8051MCU being supported
"you'll need an ICE ($500 -up)"
Exactly - you have to throw out the 89C52 just to have a bond-out chip or programmable-logic emulation connected using a pod. Even if it tends to give some advantages compared to processors with debug interfaces, the disadvantages are normally huge. And embedded equipment often doesn't have the form factor where it is possible to connect any pod.