i think i've isolated the crucial pieces of this cataloging problem, but i'm unclear how to approach rectifying it.
in membrane-based users, a special membrane-specific field, 'exact_getUserId', is used as the handle on the user, instead of getUserId (apparently to avoid getting users for which the specified id is just a prefix). in the previous released pypi versions of membrane (1.1b5) and current remember (1.1b3), the indexing of the users included this field, with its value set the same as that of the getUserId field. in the current trunk versions (as of yesterday) of membrane and remember, this field is not getting indexed, or getting indexed with an empty value, so PluggableAuthService._findUser() fails to find these users.
i've been unable to find exactly how membrane and remember are arranging to get the exact_getUserId field indexed with the right value, so i'm unable to say what the new versions are doing differently that fails to get this set.
through my explorations i do notice one big behavioral difference between the indexing of members in the old and new versions. in the old version, the .indexObject() and .reindexObject() is being called on the member objects (though the invoked methods are actually supplied from somewhere in the base BaseContent class) to do the indexing, while the methods are not being called in the new version.
it's possible - nay, likely - that there's something i missed in revising my remember trunk checkout to work with newest version of the membrane trunk. i'm going to soon create a Products.remember branch with my revisions checked in, hopefully by tomorrow sometime, so others can look at what i'm doing. i thought i'd immediately describe how i see the situation in case it rings any bells or provides enough info for others to investigate, in the meanwhile.