Like "Oh I can't be bothered to expose this correctly or use a polarizer... I know I'll just make it a HDR"
That's kind of off base IMO. I understand your point and I'm not saying that HDR is never used as a crutch.
However, by nature, failing to expose the image properly and without accurate stops of exposure in the right bracketed range, the HDR image produced from that wouldn't be an accurate HDR image at all.
And accuracy is the entire point of using HDRI. So if it's not accurate, there's no point in using HDRI.
It would result in a 'non-linear' dynamic range that would throw off all the other luminance values.
It's all about the technical aspect of measured light samples.
The premise of HDRI is to force the image into 32bit float mode so that it can represent a much higher range of luminance/radiance values than 8-16bit images.
It's the same concept as using 32 or 64bit memory addressing in your computer.
For example, the x64 architecture of new CPU's. Using 64bit addressing allows the computer to utilize much more memory than 32bit x86 architecture. Which makes it suitable for working with large datasets and CPU and RAM intensive applications like those needed for 3D rendering and animation.
The dynamic range of a digital image works in much the same way.
Assuming that once you create the HDRi mage you save it in a file format that supports 32bit float mode. Such as HDR, OpenEXR, radiance etc...
This allows the luminance values to be represented in a much higher range that traditional 8-16bit images would automatically clip.
So regardless of how well exposed a 16bit image is, it will never represent luminance values higher than is mathematically possible with 16bit range.
Simply put, a 16bit image will suffer from blown highlights and loss of detail in shadow points LONG before the same image in 32bit float mode.
A 16bit image just doesn't have enough data to retain detail beyond a certain point because it doesn't recognize mathematical values beyond that point.
The computer basically rounds off the values used to represent those higher luminance ranges. Which visually translates to clipping that lower/higher exposure range.
I hope I've explained myself well enough for people to understand.
There's plenty of information on the web which could explain it better than I can.