To good actor is people who carefully care and curate their data, just to be corrupted silently by the program.
Lets say then, keep the default but add at least some sort of "respect my data as it is, because this is a CSV with CSV extension" mode.
CSV doesn't have really data types (or even a hard and fast standard), so that doesn't really work. OTOH, Excel lets you specify data handling by column when importing from untyped text-based formats like CSV, which you should probably do if you care.
Or you should write up a standard for a CSV-like format with explicit semantics around data types (it could even be CSV-compatible so that you get no worse results than normal CSV handling if using software that is aware of CSV but not the new format) and start using it where those defined semantics are useful, and try to get Microsoft to add specific support to Excel for it.
You could encode the column names in row 1 and column types in row 2. JavaScript type names would be convenient for most CRUD work, but SQL type names might be more appropriate for more complicated data.
There's a whole lot of business use of CSV/TSV with Excel that benefits from the existing defaults. Heck, I've seen federal government websites distributing code lists that are actually in CSV format with .XLS extensions and that are expected to use the default “smart” conversion.
Even if Excel should arguably have had different defaults, the impact on other uses of changing it now would be enormous.