1.96.0

Several service alerts at once are one message.

A docker host with ten services on it goes down, and ten emails say so — then ten more when it comes back. Service alerts that fire within a window of each other are now sent as a single message naming every one of them, and their recoveries likewise:

• "3 services have no servers left", "3 services recovered", or for a mixed round "2 service alerts: 1 down, 1 recovered" — at the severity of the worst of them, with each service listed in full in the body.
• The window is 60 seconds by default, a new Combine service alerts within setting on Notifications. It composes with the grace period from 1.92.0: services that die together clear the grace together and fire in the same check, so the window only holds the message open for stragglers whose health checks trip a round later. Set it to 0 to combine only what one check found.
• A lone alert reads exactly as it always did.
• Alerts from elsewhere — a certificate renewal, a cluster warning — arriving mid-batch still send on their own rather than being folded into the services message.

The new strings are translated into all ten languages.

Upgrading from 1.95.x: nothing to do; existing installs get the 60-second default and can change it on the Notifications page.

Packages are attached below; the container image is at ghcr.io/avandeputte/haproxy-manager:1.96.0.