There are IO implementations that can optimize reading of multiple, scattered, ranges. Some example are io_uring and multihandle libcurl APIs. Currently kvikIO can only benefit from these within large reads (that get sliced into 4MB calls). When kvikIO gets a large number of small read calls, they all have to be handled separately and kvikIO cannot effectively use multihandle IO APIs.
Adding an API to kvikIO that handles reading of multiple ranges in a single call would allow optimizations in the use cases where the ranges cannot be coalesced on the caller side (e.g. highly selective Parquet reads).
TBD concrete API definition.