GitLab

The server instances are not easy to manage and in fact not very important.
We can imagine that the server instances are a pool managed implicitly.
This would help to have only one server instance per server and solve the server instance problem for connectToRequester().

We could imagine:

App app(name, endpoint)
App app(id, endpoint)

app.start()
app.connect()
app.stop();
app.waitFor();

Apps apps(name, endpoint) => the set of App
apps.kill();
apps.waitFor();

for (App app in apps) { use of key values to discriminate }

Q: Is it still interesting to be able to wait for a non-terminal state? A state handler would be more convenient. Q: However, what to do with a server that has no application instances anymore? Wait before terminating it (cf strategy for linked apps)