Page timeouts when accessing merge request (502)

I receive complains that users are unable to access specific merge requests. I am also unable to access these pages myself, they seem to timeout (502). The nginx log shows me the following errors:

2015/11/25 09:14:29 [error] 8939#0: *455824 upstream prematurely closed connection while reading response header from upstream, client: nnn.nnn.nnn.nnn, server: gitlab.servername, request: "GET /helios/helios-middleware-plugins/merge_requests/244 HTTP/1.1", upstream: "http://unix:/home/git/gitlab/tmp/sockets/gitlab.socket:/helios/helios-middleware-plugins/merge_requests/244", host: "gitlab.servername" 2015/11/25 09:16:24 [error] 8939#0: *455903 upstream prematurely closed connection while reading response header from upstream, client: nnn.nnn.nnn.nnn, server: gitlab.servername, request: "GET /helios/helios-middleware-plugins/merge_requests/244 HTTP/1.1", upstream: "http://unix:/home/git/gitlab/tmp/sockets/gitlab.socket:/helios/helios-middleware-plugins/merge_requests/244", host: "gitlab.servername" 2015/11/25 09:22:35 [error] 8939#0: *456293 upstream prematurely closed connection while reading response header from upstream, client: nnn.nnn.nnn.nnn, server: gitlab.servername, request: "GET /helios/helios-middleware-plugins/merge_requests/244 HTTP/1.1", upstream: "http://unix:/home/git/gitlab/tmp/sockets/gitlab.socket:/helios/helios-middleware-plugins/merge_requests/244", host: "gitlab.servername"

Other pages can be accessed normally. I do not spot any errors in the log files. Could someone advise how to debug what is going on?

Centos 6.7
Gitlab CE: 8.2.0