Aphyr's test is not a benchmark. It is a test of database correctness under a carefully constructed set of pathological circumstances. It cannot be used to infer real-world performance behavior.
Yes, the general advice for users of the GAE datastore is to build Entity Groups around the data for a single user. That isn't absolute though, and it doesn't cause problems for watch lists; the Watch can be part of the User's EG rather than the Issue's EG. Or it can be its own EG. In practice this doesn't require as much consideration as you probably imagine.
Yes, the general advice for users of the GAE datastore is to build Entity Groups around the data for a single user. That isn't absolute though, and it doesn't cause problems for watch lists; the Watch can be part of the User's EG rather than the Issue's EG. Or it can be its own EG. In practice this doesn't require as much consideration as you probably imagine.