when utilizing a generalized Lorentz transformation that are not constant, then one could break down an event into infinitesimal events until we get inertial(ity). Taking the positions vectors into consideration, then we could make linear transformations at each point and glue them together on a base. The arguments are, if we make a synchronized observation of a fast moving event, by changing the coordinate base constantly, there will be a contradiction on the observation compared to an stationary observer. *On a whole both these observers could be seen as inertial while considering large distances to the event which makes the slightly rotating observer an inertial observer with good approximationFor full details see:

http://en.wikipedia.org/wiki/User:Ko...ahdaei/sandbox