With apologies to
@Naka for taking this way off topic.. please don't be out off by this technical stuff which you don't really need to know.
I could be wrong here but I thought demosiacing and white balance are applied to the raw data in a single step to produce what's effectively a 16 bit per channel per pixel image.
I'm willing to accept that some raw developers may apply exposure, tone & DLR adjustments - as you call them - prior to demosaicing but I'm 90% sure most don't. If I get chance then I'll have a dig through the dcraw and RawTherapee source code, but don't hold your breath. If I'm wrong about this then everything that follows will be incorrect too but all the evidence I've seen points my way.
From then on the data is no longer raw. Every manipulation - including applying tone curves, exposure adjustments, or modifying the black and white points are applied
after this point and can only degrade the image data. You might be able to stretch the histogram and if you're careful do some interpolation, but you won't be getting more out of the data. The exposure slider is in some sense safer than moving black and white points because it does it's best NOT to stretch the histogram and introduce the familiar 'comb' appearance.
Some raw developers apply gamma curves before the final rounding to 16 bits but most don't, and there is some research that it makes no effective difference - but that only really applies at the final output step anyway and has little bearing on the case where any of the other sliders are being tweaked.
The only way to effectively increase the available information is to expose to the right without clipping and reduce the exposure in post; that can effectively increase the amount of detail in shadow areas. Even then, though, it's reasonably well established that reducing ISO where possible is a better option than ETTR; it's only useful when you're at base ISO.