Just wanted to pick up on this minor point: > current bottleneck seems to be RAM RAM exhaustion is often a secondary side effect of the true underlying bottle

Just wanted to pick up on this minor point: > current bottleneck seems to be RAM RAM exhaustion is often a secondary side effect of the true underlying bottleneck. It could be that RAM used for temporarily processing an HTTP request, say, would be freed quicker if the database returned sooner. Maybe the database is the actual bottleneck. And the actual root cause is an indexing / partitioning issue.