Hey guys, feel like I am going crazy trying to bring spicedb into a production setting I have had some ridiculously bad instances of my authzed instance not being able to resolve in 10 seconds, which was pretty crazy since I think authzed is technically scalable to 5ms? Testing it separately using the zed CLI against the production load showed me that it is able to resolve in under 20ms max. The characteristics are: 1. I am currently using experimental service BulkCheckPermission - Does it matter if I switch it to the stable API or not? 2. I am able to get traces to flow into datadog normally but on those random performance issues calls (on the top left, that authzed API call took 9 seconds), and the traces don't even reach datadog. The screenshot also shows what happens normally when the traces reaches correctly (on the right) 3. DB metrics are entirely stable, no deadlocks, well below max connections (which is at 401, I am using db.t3.medium, postgres 14.9) 4. The problematic requests don't always take so long. In fact, they usually resolve just fine. 5. I am using the authzed/authzed-node library as client. version 0.13.0 😅 6. No CPU or mem issue on the spicedb nodes either. if anyone has ideas on what might be the issue, I would really appreciate if you holler at me! https://cdn.discordapp.com/attachments/844600078948630559/1356899291913916497/image.png?ex=67ee3e71&is=67ececf1&hm=1d2eb9ef51dab26a98db2fdc3c67c339cb0699cc48db56155a5d244640a27d69& https://cdn.discordapp.com/attachments/844600078948630559/1356899292383543446/image.png?ex=67ee3e71&is=67ececf1&hm=6ecdc1b4585271ecbfb156d59eb62b9c192f4035c5c7726773a62b82bc8e8a80&