With a 32-bit ISA already in use, the continued use of SPARC makes much more sense.
In addition, moving the ISA decision back several years also removes some of the reasons that SPARC would have been a less reasonable choice. (It seems that the program leading to the ERC32 began no later than 1991. At that time, the continued existence of SPARC workstations would have seemed very reasonable, there would have been greater hope that the openness of SPARC would have broadened adoption, and a custom ISA might have seemed a less reasonable alternative.)
Editing the question to ask about the choice of SPARC for ERC32 makes the question substantially less interesting--the choice of SPARC at that point seems much more reasonable. Likewise, editing the question to include the "new" information about the previous use of SPARC while still asking about LEON makes for a less interesting question--even with portable software an ISA change would add significant cost and risk. Posting the "new" information in an answer (i.e., answering the question as it stands) might not be unreasonable; such an answer should include something of the rationale for SPARC for the ERC32. (If I posted such an answer, I would be tempted to make it community wiki since I know it would not be a complete answer; but it is not clear if that would discourage other contributions.)
I am disinclined to close or delete the question. Even though it is not as interesting a question as I thought, it is still (I believe) a valid question and it has one favorite, 4 upvotes, and 1 downvote.
Even though I am asking about a specific question, addressing the more general question of the title as well might be more useful than just addressing the specific question.