12M sound bit weird number for limit, I wonder where it stems from. Similarly I wonder why row count is limited at all, or at least why the limit is not something huge like 2^32 (or available RAM).
For various internal indexing purposes, it's more convenient to use powers of 2 or their multiplicities. While "no limits" is doable it would have some impact on the performance and would require a bit more complex UI for the column-oriented editing actions (which might not necessarily be worth it).
BTW, data tables exceeding 12 million rows can be handled by a companion database program GS-Base ( https://citadel5.com/gs-base.htm ). It's a database with spreadsheet functions. Uses up to 256 million records/rows including data types same as in GS-Calc and binary fields: Long Text, Images/Files, Code (for code snippets with syntax highlighting for 16 languages). It uses the same type of calculations within one and more tables but in this case, they record/row based.