The CDB_EstimateRowCount function executes an ANALYZE in the affected tables of the query when that tables don't have stats. But this operation could come at a great cost if, for example, the table is big so we should add an optional parameter to skip the analyze part where performance is more important and we can deal with tables without stats.
This issue arises because in the tiler we're using this function to gather stats from the layer queries and execute an ANALYZE could penalize the performance