Investigation of observed different caching between spymemcached and moxi in one environment shows that moxi is inadvertantly using the default from libmemcached. I say inadvertantly because that default may actually be the less common choice. It is different than what Enyim, Spymemcached and libketama do by default when selecting Ketama.
libmemcached seems to use non-weighted Ketama, which seems to be specific to libmemcached and does not define the hash continuum the same way. Specifically, it seems to define the continuum with 100 points per server, rather than the 160 commonly used.