Phantomical
The implementation currently reads from the start of the file and discards data before the requested offset. We should ideally fix this, or at least fall back to `AsyncReadManager` for offsets > 1KB. This was originally disabled due to some reported issues that roughly looked like Seek not working correctly, it's possible a closer look at the APIs would let us go back to using seek correctly.