- Shipped
- September 23, 2026 at 7:28 AM UTC
- Author
- Kamo
- Commit
- 9f60d25
Hosted Computers was reachable by both automatic provisioning paths, and either one would have put it in front of every organization on the platform at once. FeatureController.list auto-provisions every available non-core app that has no OrgFeature row, on a plain Settings page load. **************** grants an app the plan matrix never mentions when kamo.entitlement.fail-open is on. Neither is wrong for a screen; both are wrong for a KubeVirt VM that reserves 2-16 GiB of a node's memory and tens of gigabytes of Longhorn, because the scheduler bounds the fleet by REQUESTS rather than usage - so capacity is spent when a computer is assigned, not when somebody uses it. The organizations that reached the app first would have exhausted the nodes for the ones that were deliberately given it. NEVER_AUTOMATIC already closed exactly these two paths for the EHR, and its note says the next app of this kind should be one entry. This is that entry, added for the other reason an app must be asked for: not a compliance boundary, a finite physical resource. It is also what makes the platform Apps & Features switch mean something here - an operator turns it on per organization, which is what "deliberately" is defined as in this file. The set stays narrow, and the test that pins it narrow now says why each of the two members is in it.