I usually side with the circumventers on this one. This misbehavior happens for a reason, which usually is that it's impossible or infeasible to do the work with "proper" systems. Excel sheets on pendrives is what you get when people can't exchange data using "correct" software, or when people need to continuously iterate on the shape of the data set, while getting the schema changed in the "correct" system would involve a ticket to IT and 2 months of waiting.
I think instead of paying people for doing a job and then doing everything possible to make that job difficult, companies should embrace that people will work around any deficiencies of top-down systems, whether software or procedural. But I guess this tug-of-war exists since the time of first corporations.
And then there are examples like the production planning done in Excel because SAP PP isn't just good enough. Then the guy who created said Excel tool left. Years later when production had to switch to weekend shifts zhey couldn't because no one could adopt the Excel tool. Once you reach that point you are screwed.
But there will always be a use for Excel as a tool for informal data-monkeying and analysis - exporting a table and doing some pivots to summarize or use a different view. Non-technical users will always expect this capability and flock back to it. Not everyone's a programmer or has the interest to be - you'll have to pry excel out of most accountant's cold, dead hands for example
Issue I have with that approach is not using excel per se. It is using Eycel in addition to what ever system is being used the first place (SAP for example is a pretty popular thing to circumvent with Excel sheets). That and over engineering spreadsheets to the point where nobody but the creator can use them anymore. At that point you data just gets suspect. And circumventing existing systems with local offline spreadsheets just screws up everything. Now combine these two.
But that has less to do with spreadsheets themselves, see my fist point regarding cross docking, and more with the application. Used correctly spreadsheets can be incredibly powerful. And at least for my purposes I have yet to encounter an analysis issue I failed to dig through using Excel. Sure, something like Python might have worked better but consider me to be an empowered user (read: I have no clue about Python or SQL or...). And at least I knew I could trust the way the analysis was done as intended.