Hi Eliot, I tried out your new 4.0-3018 today and wanted to report my experience.   A litmus test whether I can bring it into my dev environment is if the Magma suit can pass.  If it can't, I can't use it heavily yet.

I started with your converted trunk46-spur.image and loaded Magma.  Something strange happened while loading, LanguageEnvironment class>>#localeID: got into an infinite recursion because the KnownEnvironments dictionary (class-var) got a bunch of new entries into it, but left needing to be rehashed.

I don't know if this is something related to the trunk46-spur image or something else.  At any rate, I was able to get past it by changing that method to rehash the knownEnvironment dictionary before recursing, thusly:

     ^ self knownEnvironments
          at: localeID
          ifAbsent:
               [ self knownEnvironments rehash.  "<--- added this"
               self localeID: (LocaleID isoLanguage: 'en') ]

After witnessing a glimpse of what appeared to be _exhilerating_ performance for the load, running the test suite got about 20% through before one of the client images crashed.  I began the process of inserting halt and restarting the suite to see if I could zero in on a specific operation but then it crashed slightly earlier.  Okay, halt just before there -- stepping through, everything looking fine, about to do a commit.  Boom.

There is a lot to the commit process but before trying to debug it further, I'd like to at least say, the ".log" file of the client that crashed is reporting the crash originating out of exactly the same area that original Cog was crashing many moons ago (but which you have had fixed some time ago and never saw it at all in 2776).

I backed off to a much simpler test suite, the MaObjectSerializer suite.  It passed.  _Blazing_ fast too.  Awesome stuff, can't wait..