Could you adapt the idea od fork/join parallelism in something like Verilog to allow cell-level parallel processing?
If I had cell1 that got data from web-API 1; and the next cell, cell2 got web data for web-API2 then the idea is to add a way to of stating that cell1 and cell2 are to be forked-off, (and their outputs updated independently), whilst any subsequent cells after them are not executed untill cell1 and cell2 both finish, (the join).
Systemverilog adds other abilities to its fork/join that shouls maybe have analogues for jupyter cells, but the idea is to allow straight-forward to express parallel execution at the cell, (and multi-cell), level.
Whatdoyouthink?