Comment 3 for bug 1677434

While working on fixing bug #1454661, I added some instrumentation.

It seems that doing "juju models" is creating state objects for every model, and loading all of the presence information.
While doing debug information, if I run "juju models" in one window, while logging is enabled in the other, I see:
machine-4: 19:06:12 DEBUG juju.state.presence looked up 440 unknown sequences for "c3525263-1fe3-4556-8b5c-2bbc3beee6f7" (6 unowned) in 13.677754ms (32.2seq/ms) from "presence.beings"

The "440 unknown sequences" is code for "I had to lookup the presence information of 440 values in the database that I haven't seen before".

As these are machines that are running "stable" they should have already cached all of the presence information for all of their models.

I can trigger this reliably with 'juju models', but I *cannot* trigger this with 'juju status'. So Status appears to be using StatePool and thus being able to give information from what it has already cached in memory, while 'juju models' is reloading that information from mongo.