I ... just prototyped the spreadsheet hackery required to count invalid values in a column of CSVs, and report the first offending row number ... all in a one-cell formula.
The hackery : three to four bits of code are completely redundant. The missing feature here is the explicit assignment of intermediate results to a variable, allowing in-cell code-reuse ( the intermediate results are probably already re-used, under the hood, implicitly ... I hope ).
So, if spreadsheet formula semantics included assignment operations, we would probably see an uptick in spreadsheet utilisation, and a reduction in code length across the industry.
Maybe the feature is banned because it would just make spreadsheets even more complex than they are already. But I would see it as a marginal improvement overall.