COVID does not constitute a natural experiment!
A useful natural experiment is when something changes in a way that can only (or at least almost entirely) affect your outcome by way of the cause you're interested in. In the cause and effect chain of A -> B -> C, if A affects B, but not C, then you can exploit wild things in A to learn about how B affects C. But if A affects C outside of B, you can't tell the effects apart.
COVID does not constitute a natural experiment!
It's a mantra that apparently needs repeating. The pandemic changed work from home. Work from home presumable changes productivity. The pandemic also changes productivity, so you can't tell where the pandemic-driven productivity changes are from more granularly. Learning anything that generalizes requires tremendous care.