The health page now reports what the portal measured — and what it did not

A portal's health page lists whatever is currently wrong with it. That is the right thing for a verdict, but it made two useful numbers impossible to read: how much of the portal's content was already prepared when it started, and whether it managed to gather that content in one piece. Both were written to the startup log and nowhere else, which meant that answering a question about them needed access to the log store rather than a browser.

Both now appear on the health page, and they are reachable without signing in.

What the portal's own preparation found. When a portal starts, it checks which of its content types already have a prepared, ready-to-use build and which it will have to build itself. The health page now carries that result: how many types it looked at, how many were ready, how many were pending, and how many it had to reclassify because its own preparation had moved on faster than the list it was working from. The count of already-prepared items is printed beside it, together with a note that the two numbers count different things — a pair that has been misread as a contradiction more than once.

Whether the content was gathered in one piece. When a portal does have to build content, it first collects the source for everything at once. That collection finishes when the answers stop arriving, so an unusually long pause in the middle can end it early and hand the builder an incomplete set. The health page now reports how many batches of answers arrived, where the collection settled, and the longest pause between batches measured against the window that ends it. A long pause is flagged; a short one is stated plainly, which is just as informative.

A clean reading is printed, not silent. Both entries print whatever they find. That is deliberate: a reading that appeared only when something was wrong would be indistinguishable from a portal that never took the reading at all, which is exactly the ambiguity these two numbers were stuck in. So "this portal has no reading" and "this portal has a reading and it is clean" are now two different sentences on the page, and neither is silence. A portal that never produced a preparation result at all says so and is marked as needing attention — an absent measurement is never reported as a clean one.

Nothing here changes whether a portal is considered healthy enough to serve traffic, or to be restarted: these readings are published for people to read, and they carry no authority over either decision.