1: Those bounds checks come at a cost of having overflows getting in the way of actually checking bounds. An overflow of an unsigned value results in just a small number... that means that you still have to check the lower bounds no matter what.
2: They store the same number of values as unsigned, it's just handling of those values and comparison operations differ. A signed and unsigned 32bit integers store 2^32 values.
3: Crypto algorithms do not operate numerically, thus 2147483647 * 100 will yield exactly the same result for uint32 and int32 as much as any crypto algorithm is concerned.
4: Therefore you send 2147483648 in binary form as 4 bytes = 0x80000000