Feature: support resource sharing for sharded logs #589
Description
As part of #574 I was documenting how to shard logs. The current solution is simply to bring up a whole new stack, including DB instances, buckets, VMs, binaries, load balancing rules, etc.
At least some of this could be reused. e.g. in GCP, a single Spanner instance could host multiple databases (one for each log). #267 covers this aspect. It's also logically reasonable that a single personality could route through to multiple log instances. Technically this is possible at the moment, but the monitoring being added as part of #570 will not disambiguate which log the counters. For some of the gauges that are exported, running multiple log appenders in the same process could totally break the monitoring dashboards as the two appender monitoring tasks stomp on each other.
This issue is raised to allow us to collect a demand signal, and ideas from the community on the features expected.