We realized that there are certain usage patterns which would cause Extension:JADE to create unacceptable numbers of pages. Since we create a JADE-namespace page for each revision being judged, we cannot allow a judgment for every revision. This behavior would quickly swell the page table to many times its current size, and would at least double the revision table size.
Some quick estimates (detailed in comments below) show that creating a JADE entry for rule-based, automated events such as assigning an "autopatrolled" flag is an anti-pattern that we must discourage at this stage. The volume of these patrolling tags can quite easily equal the number of revisions being created. This has already become a problem for the logging table (T184485: Stop logging autopatrol actions) so autopatrolled events are being discarded there as of a few months ago.
Expected usage at human scales will eventually result in about 0.5M additional new pages created per year, if all existing workflows can be migrated to create judgments.
Our working conclusion is that we need to rely on social agreements to not do silly things with JADE until the technical limitations are overcome. There is strong precedent for this approach, for example https://en.wikipedia.org/wiki/Wikipedia:Bot_policy which already covers the situation we're looking at. Generally, bots are not allowed to create millions of unneeded pages and that's exactly the situation that we're concerned about here. The remaining work would be for our team to write a clear statement about what is currently harmless vs harmful JADE usage, and circulate to the community that might integrate tools with JADE or otherwise use the new namespace.
In the long run, we do hope to overcome the storage limitations, so that we can allow bots and other interesting information sources to populate JADE in ways that make sense, in addition to the humans that we've had in mind as our primary users.
Current thoughts about scalability are being summarized here,