Currently we use `estimatedCount` to fetch the document count for time-series collections. Currently this fails for these collections (screenshot of error attached). We should probably just fail silently for these collections if we cannot fetch the count without doing a full scan.