I feel like this is one of those bugs that would have taken less time to fix than it took to document.
After many years, I still have to slap myself when I catch myself thinking this way. Unless you know the architecture of the software, I've learned that it's often significantly harder than you imagine to fix seemingly simple and obvious bugs without breaking something else.
Fine, a slight correction then. It should be easy to fix except for terrible software architecture. Terrible software architecture is not a valid excuse when you have plenty of time and billions of dollars. So there should be little forgiveness for this bug still existing.
As a consumer affected by this bug, sure, you may find it inexcusable. But the basic truth about any mature software product used by tens of millions is that there will be a laundry list of bugs competing for limited resources. So there will always be bugs that you want to fix but simply isn't a high enough priority relative to all the other bugs or features.
There are limited resources, but there are enough resources to improve the architecture over more than a decade. And if they decided to avoid architecture changes, there would be nothing in the way of throwing lots of siloed software engineers at bugs to get more of them fixed. They're not lacking for money with so many customers. It's possible for them to set a much more inclusive cutoff.
They could have fewer bugs per feature than almost any other product after being stable and popular for so long. But they choose not to make that a goal.
This is basically what happens, at least on my Mac - it complains about it not being a valid SYLK file, then I think, "I have no idea what that means", then I dismiss the error and it opens it up as a CSV.
You could even do something as trivial as "if Excel fails to open the file as SYLK, try again as CSV" and cut at least 99% of the problem away.