Managing side-effects of e2e tests

[Nachasic]

E2E tests perform a lot of operations with the backend: creating, updating and deleting spaces, updating user profiles as well.

When same tests are ran against staging multiple times simultaneously (e.g. in parallel jobs) this creates data races. When testing anything resolved by name, like spaces, these data races can be mitigated by incorporating RNG for names. However, when it comes to testing invitations such strategy wouldn't work.

Scheduling test jobs to run them sequentially doesn't seem feasible, so sandboxing the environment comes to mind as the most sensible solution.

Alternatively, this problem could be solved by allowing the creation of users - each test suite would create it's own set of users, using a naming convention and RNG for emails, effectively limiting the scope of test operations to a set of short-lived user accounts.