In the RISC-V vector extension, instructions are provided to load/store single or multiple vector registers in a single command.
Among these, there are currently four types of load instructions available from memory to v registers, namely vl1re8/16/32/64.v.
However, these four types of instructions share a common range to load, which is determined by the "VLEN bit". For instance, the vl1re8.v instruction stores VLEN/8 [byte] into the vector register, and the vl1re64.v instruction stores VLEN/64[dward] into the register. In other words, the size loaded into the vector register is the same for all four types of instructions.
When VLEN is less than 64, there may be instructions that cannot be used. However, when VLEN is 65 or greater, the results obtained from using any instruction should be essentially the same. The only conceivable reason for this redundancy might be if there were alignment restrictions for loading from memory.
The RVV spec provides an answer to your question:
The alignment requirements are also different, the memory needs to be aligned to the EEW.