Uptime: Wiki | v0.1.24 (Early Access)

Fixed

Pressing Escape while renaming a rack now closes the name box properly. Escape cleared what you had typed but left the box holding on to the keyboard, so from then on nothing responded: moving, interacting, the pause menu, all of it went into the text field you thought you had left, until you clicked somewhere else to break it out. Enter had always released it correctly. Escape never did.
Escape now puts the rack's original name back. Once you had started typing there was no way to change your mind. The field kept whatever you had entered and the only way out was to accept it, so a rename you opened by accident had to be finished. Escape now restores the name the rack had when you opened the field, so cancelling a rename actually cancels it.

Putting a server into maintenance now tells you what it could not move. It reported a bare count, so "1 can't move yet" left you hunting for which workload was stuck and why, on a server that looked otherwise empty. It now names them, the same way the message does when you try to lift a server out of the rack: the customer, the kind of workload, and how much storage it is holding.
The migrate list no longer offers servers your workload cannot run on. A customer whose software is built for one CPU architecture can only be moved onto servers of that architecture, and reserved machines are held for work that asks for them by name. The same goes for accelerators: work that needs a GPU only runs on a GPU machine, and work that does not need one is kept off them so your GPU machines stay free for the customers paying for them. The list was ignoring all of this, so it would show you an incompatible server as a perfectly good target, let you pick it, and only then refuse the move. Those servers are now left out.
The migrate list is sorted by how much room a server really has. It ordered by free memory alone, so a server with plenty of memory but almost no processor headroom sat at the top of the list. It now ranks on whichever of processor, memory or storage is tightest for the workload you are moving, so the first option is the one with the most genuine room for it.

A widely-felt incident no longer stretches the incident page sideways. Every affected tenant on a machine was listed by name on that machine's row, so an incident touching a couple of dozen customers wrote a line long enough to widen the column, shove the rest of the page out of view and put a scrollbar under the whole panel. Each machine now names the first couple of tenants and says how many more it carries. The full list is unchanged in Affected tenants directly below, where each name is still a link.

The wall display no longer shows a heading that disagrees with the board. The title above it could name a different view than the one you were looking at, usually falling back to Security whatever was on screen. It happened after loading a save, and again every time you pressed Tab to switch between plain and engineering wording. The heading and the board now always match.