24-bit packed pixel modes are a bit of pain to support, but without knowing how the rendering is carried out, it's a bit difficult to say how feasible it is to fix.
An indirect rendering strategy that renders to a 32-bit BitMap in fast ram that is then copied to any freshly exposed or updated bit using BlitBitMap() might work since the bottleneck is likely to be transferring the data over the bus rather than any conversion overhead. However, since I don't know how the browsers rendering engine works, that might be a completely stupid suggestion.