I suspect that it would be possible to insert a new layer of logic an memory between the original 32SII chip and its I/O (keyboard and LCD).
Some currently unrecognized keystroke combination would bring the new environment to the fore, disconnecting the original chip temporarily. The new layer would replicate the 32sII program editor, but enhance it with named program storage and (perhaps) named register arrays.
New commands would cause the loading of a named program into the actual 32sII's program memory by keystroke simulations. Likewise, resulting data (the 27 registers) could be copied out to the new memory as an array of values, perhaps to be reloaded later.
The new electronics could include fast stream I/O capabilites and could be supported by PC links and software. The physical ciruit board might be installed in place of the original, with the 32sII board then attached as a daughter board to the new one.
Taking it a step further, one might replace 32sII electronics and logic entirely, with a new, more capable implementation.
Buutttt . . .
H-P appears ready to solve at least some of the 32sII's problems with the 33s (albeit, while introducing others!), so it would seem to make sense to wait and see just how that unit behaves before embarking on an elaborate upgrade for a discontinued model. Another approach might be to get one of the other new models (48GII or 49G+) and obtain or develop a four-level stack emulation for that environment.