As seen from HELP-91015, it is not at all clear from default logs whether a pipeline is doing a local read or remote read. This information would really aid debuggability of these scenarios.

There is a level 3 DEBUG log for when a shard does a local read for a pipeline here. We could consider promoting this to default verbosity and creating a similar log to indicate if it is dispatching remote read.

Another approach would be to increment counters that indicate reason for a remote read or a local read, e.g. shard topology out of date