direct indexing api relies on a single solr client instance, which gets reconfigured on the expiration of the cached info retrieved from the IS, indicating the solr collection to be used
reintegrated branch solr75 -r53793:HEAD
catch and ignore all the exceptions to keep the autocommit task scheduled
internal calls are not cached, so the cache management has been extracted to dedicated class
Cache for indexDsInfo
index clients reuse the connection. Using Guava's AbstractScheduledService for the commit task