I believe that tables in Org are already too complicated due to the
spreadsheet feature. However those who are brave enough to add cells
spanning columns and rows may take some inspiration from
reStructuredText, in particular horizontal lines marked by "+===+===+"
https://docutils.sourceforge.io/docs/ref/rst/restructuredtext.html#tables
Grid tables are described with a visual grid made up of the characters "-", "=", "|", and "+". The hyphen
("-") is used for horizontal lines (row separators). The equals sign ("=") may be used to separate optional header rows from the table body
(not supported by the Emacs table mode). The vertical bar ("|") is used for vertical lines (column separators). The plus sign ("+") is used
for intersections of horizontal and vertical lines. Example:
+------------------------+------------+----------+----------+
| Header row, column 1 | Header 2 | Header 3 | Header 4 |
| (header rows optional) | | | |
+========================+============+==========+==========+
| body row 1, column 1 | column 2 | column 3 | column 4 |
+------------------------+------------+----------+----------+
| body row 2 | Cells may span columns. |
+------------------------+------------+---------------------+
| body row 3 | Cells may | - Table cells |
+------------------------+ span rows. | - contain |
| body row 4 | | - body elements. |
+------------------------+------------+---------------------+
RST tables have a feature making hard to implement autoalignment. If
come cell have "|" with offset in respect to lines above and below then
it is a regular character.
Besides "grid" tables there are "simple" tables without vertical lines.
Perhaps it better to implement new table features as src blocks for some
new language and a dedicated Emacs mode. In the case of success such
proof of concept may be merged into Org core.
My impression is that Org tables quickly become hardly maintainable when
their complexity is above some quite low threshold. E.g. automatic
recalculation works only for first #+tblfm: line. It requires some
efforts to figure out association of particular formula with cell spans.
Merge cells add more complexity to formula ranges. Some protocol should
be defined to allow source blocks to generate extended cells.