The reduction in the "keyed nodes" on the Stats side is an unfortunate side-effect of some necessary rate-limiting we had to do on all statposts from nodes. Before the server upgrade, the server the stats system was running on was overloaded causing other problems. We broke apart the services onto multiple servers, designating one as solely for processing the stats.allstarlink.org information. With the overload condition fixed, we discovered that the overloaded system was acting as a "natural" rate-limiter. To actually process all the keyed-up posts, the cost to run the service was not worth the value of the data, especially as the stats service isn't necessary for ASL operations and the vast usage of it is to check node online status and the bubble charts. The way it's designed today, all statpost calls must arrive at the same server. In cloud architecture, adding cores that are "always consumed" get pricey very quickly. So, for now, a statpost is only processed once per 30s per node. It's very safe to say that the Statpost service's load from the huge number of nodes has far, far exceeded what the architecture can handle reasonably.
We're looking at a complete redo of the architecture so we can have more smaller systems process statposts and probably some other system display the web-visible information. We're only in the early staging of working out exactly what we want to really do at the scale we are now. Perhaps we'll hide all the "keyed up" information until it's more reliable again.