I don't know which, if any, of these bugs you want to address in any
stable releases.
If fixing bug 276 "rte_rand() bit 31 and 63 are always zero" is enough,
then one could either split this patch set in two (w/ only the new
rte_rand() going into stable), or develop a smaller fix based on lrand48().
Since the old implementation was an inline function, any fix will
require a recompilation.