After the babl space invasion, Multiply gives wrong results
As can be seen in the screenshot below, Multiply blend mode gives different results in GIMP-2.99 post-babl-invasion, compared to GIMP-2.10. GIMP-2.10 results are correct. Multiply is a chromaticity-dependent editing operation and so is expected to give different results in different linear gamma RGB working spaces.
To make an easy-to-see example, I chose two colors, one cyan and one magenta (not RGB cyan and magenta, but rather according to LCh color wheel), such that in the linear gamma Rec.2020 color space multiplying the two colors together produces an almost neutral gray.
In GIMP-2.10, converting the linear Rec.2020 image to linear built-in sRGB produces an immediate color change, as it should: instead of nearly neutral the resulting color is a fairly pronounced cyan-blue.
In GIMP-2.99, the result of multiplying the two layers in the linear gamma Rec.2020 color space produces the exact same color as produced in GIMP-2.10 after the image is converted from linear gamma Rec.2020 to built-in linear sRGB. This is wrong, as if Multiply were really being done in built-in sRGB:
This wrong behavior for Multiply in GIMP-2.99 - that multiply in linear Rec.2020 and in linear built-in sRGB produce the same results - has immediate editing implications, which are born out by experimenting: White balancing using Levels or by dividing by the color cast that you want to remove, both fail in GIMP-2.99 unless the color cast was produced in the linear gamma sRGB color space instead of some other linear gamma color space.
Here's a copy of the XCF file: multipy.xcf