TIFF is the standard format for scientific imaging. It's the container format used by well over 80% of all imaging formats out there.
It's being slowly replaced by HDF5 for some applications.
But all of the features you mention above are tunable and are in practice perfectly fine with TIFF containers.
The big problem with highly tuned formats is that they make a lot of choices for the user. This works well when the format is used exactly as designed, but that is often not the case. With a tif I can pick exactly the compression, bit depth, color mode, tiling etc. that I need for the specific data.
> All those features can exist for any lossless image format.
Yes, but for most image formats they aren't defined. They are for TIFF.
The problem, of course, is that if you use an exotic TIFF feature, very few TIFF readers will understand you.
1. On-disk size
2. Compression/decompression speed
3. Access speed (for use in image analysis algorithms, editing, compositing, etc)
4. GPU friendliness (which operates in warps on small blocks of contiguous pixels in image space)
TIFF doesn't really optimize for any of these (not even #2 since in-memory decompression can be faster than paging uncompressed data from disk)