The problem with HTTP 503 Server Unavailable seems to occur intermittently.
In the French datacenter Invantive Cloud now uses a load balancer to distribute and direct the load across nodes. The load balancer uses a health check which can be imitated using curl (see How do I use curl to execute an Invantive Cloud OData4 request?):
curl -sS -k -H "Host: bridge-online.invantive.com" https://<backend node>:<backend port>/healthtest
However, the SSL-connection can not always be established due the error:
curl: (35) OpenSSL SSL_connect: SSL_ERROR_SYSCALL in connection to <backend node>:<backend port>
This causes the load balancer to consider the backend node to be failed. A few seconds later, the backend node on the next pool is considered healthy again and so on.
An analysis is ongoing what causes the health monitoring to fail intermittently.