Without having been involved with the testing its pretty difficult to know for sure, but it took SN a number of releases to get the R972 functioning correctly and based upon comments and reviews it appears to have still had some issues. One can easily envision a scenario where more memory was required for the Trinnov processing and that in a hardware design with a fixed, unexpandable amount of memory, this lead to design trade-offs that impacted other parts of the stack. Optimizing and rewriting code to free up memory(perhaps at the assembler level) is never easy and requires really talented coders.