Ah, thanks for that. That’s interesting, as it’s not correct (or at least no longer), as my post highlighted – bitmapData is not an “explicitly mutating method” neither in name, intuition, nor documentation, yet it is not thread-safe, contrary to what those release notes say.

That section below that, though, is helpful (“NSBitmapImageRep: CoreGraphics impedance matching and performance notes”). It confirms what I’d suspected based on disassembly and backtraces, that NSBitmapImageRep is usually a thin wrapper over a CGImage.

And I suppose you could try to read between the lines – taking both those sections in mind – to deduce that maybe bitmapData et al are in fact mutating as an implementation detail because apparently they require copying the data out of the CGImage (but apparently in a way not privy to the thread-safety that other mutations are, like caching scaled & format-converted versions for screen rendering).

I know that Apple recommend just drawing the bitmap image to a bitmap context with known pixel format, rather than using bitmapData, although I’m not convinced that’s the most efficient way in some cases – I haven’t reverse-engineered or benchmarked it exhaustively, but I do that sometimes in one of my current projects, and it’s slower. 🤔