A strategy patch, and a page add, delete or reorder, is checked against
the strategy's revision and moves it on when it lands. A change waiting
behind one of this device's own that was being sent still expected the
old revision, so the server refused it once the first landed: a page
reorder made while a map change was sending went to attention, with no
teammate involved. Keeping a waiting patch's revision through a merge
(the previous commit) made a map and theme change behind a reorder hit
the same refusal.

When the server accepts one of this device's own strategy-revision
changes, moving the strategy from r - 1 to r, every waiting change that
expects r - 1 now expects r. Only an accepted answer moves waiting work
on: a revision read live says nothing about who made it, and a
teammate's leaves the work to be refused and the user asked.

The test server now checks page adds, deletes and reorders against the
strategy's revision and chains a batch's ops made from one revision, as
the real server does, and can hold a batch's answers after the live read
shows it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>