latency – What is ASIO input and output buffer? Why are they typically the same?

Why separate input and output ASIO buffers are needed? And why both buffers are typically the same (e.g. 128 samples each or 256 samples each, etc)?

I imagine that the function of the input buffer is to accommodate any hiccups from the OS before getting audio from the interface to the host software. It is the output buffer where the heavy lifting occurs: virtual instruments and effects – on top of the OS hiccups.

I would expect the output buffer needs to be larger: typically processing larger chunks of data is more efficient. Why isn’t it the case? Is my understanding of the buffers incorrect? Or are the unpredictable delays from the OS the primary reason the buffers are needed?