We are experiencing intermittent, severe delays when loading the message list in the SOGo web interface.
To diagnose this, we configured extended logging in Nginx.
The metrics reveal a significant discrepancy between the time Nginx spends waiting for the backend and the actual processing time reported by SOGo.

Log Analysis Example:
Nginx Log:
email.xxx.xx - xx.xx.xx.26 - - [14/May/2026:17:00:17 +0300] "POST /SOGo/so/otgruzka@xxx.xx/Mail/0/folderINBOX/changes HTTP/2.0" 200 27 "https://email.hoster.by/SOGo/so/otgruzka@xxx.xx/Mail/view" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/148.0.0.0 Safari/537.36" 2.149 uсt=0.000 uht=2.144 urt=2.144
SOGo Log:
May 14 17:00:17 sogod [3636941]: xx.xx.xx.26 "POST /SOGo/so/otgruzka@xxx.xx/Mail/0/folderINBOX/changes HTTP/1.0" 200 27/152 0.182 - - 0 - 17

As shown above, SOGo processed the request in 0.182 seconds, but Nginx recorded an upstream response time (uht/urt) of 2.144 seconds.
This leaves an unaccounted-backlog delay of 1.962 seconds (2.144 - 0.182).
In worst-case scenarios, the browser waits for a response for up to 30 seconds.

Questions:
Where is the request spending the remaining 0000015:0000002 seconds before or after SOGo processes it?
Could this indicate that requests are getting queued in WOListenQueueSize before hitting a worker, or is it an issue with response buffering in SOGo?
What adjustments or optimizations can we implement to resolve this latency bottleneck?