regression: building disappears

[polarbearing]

This building http://www.openstreetmap.org/way/341791043/history used to be rendered correctly, and even when its disappearence was reported on the tagging list, I could load a page with correct rendering. However, zooming and refreshing caused the building disappear for me as well.

First suspicions are a regression in carto for the combination of amenity=school and building=school, or a glitch in the rendering database?

Cached or different tileserver? initial view:

Refreshed view:

Tile servers say it was rendered Thu Apr 13 15:30:06 2017
http://b.tile.openstreetmap.org/18/63088/94343.png/status (all, a b and c), the tile showing without the building.

This was the question on tagging.

[polarbearing]

Hm, I probed a couple of other amenity=school + building=school, with and without campus grounds, they all render correctly.

[mboeringa]

I see it properly displayed at all scales from Z13 onward where buildings start to appear, even after hitting F5 several times. I am in Europe by the way... don't know if there is any geographical distribution of tile server load.

[polarbearing]

So how do you see the single tile cited above? This it what I get currently, the tip of the car park, and no building:

I wonder if @tomhughes could help here and check rendering/caching machines.
Can we compare what http://tile.openstreetmap.org/cgi-bin/debug delivers, that might help him, for me:

tile.openstreetmap.org debug
Server Stats
Cache Server: kalessin.openstreetmap.org
Render Server: scorch.openstreetmap.org
Load Average: 8.42

[mboeringa]

So how do you see the single tile cited above?

Z19 view (but as said, I see the building on all other zoom levels as well):

[mboeringa]

Can we compare what http://tile.openstreetmap.org/cgi-bin/debug delivers, that might help him, for me:

tile.openstreetmap.org debug
Server Stats

Cache Server: trogdor.openstreetmap.org

Render Server: scorch.openstreetmap.org
Load Average: 8.03

So the only difference is the cache server, trogdor vs kalessin

[pnorman]

Based on the above it doesn't seem to be a style issue.

[polarbearing]

Tom has identified that as a rendering database glitch, see the operations (156) ticket above.