If your time enabled sensor (the one using GPS to synchronize) had a master ANT channel broadcasting that time value, the other sensors could pick it up with approximately one message period delay.
My reasoning for these timings is that if the master wrote its current internal time to the ANT chip (in the form of a broadcast message) as soon as it got an EVENT_TX, then that message would get sent one message period later. You could even take the internal time and add one message period worth of time to it so the slave devices could just use the value.
Any timing error would come from the time between the EVENT_TX coming across the UART, and the MPU processing it. This should be very minimal (less than 10ms).
It also has the added benefit that the ANT message period corresponds to update latency, and not accuracy. So you can potentially set your period to 0.5Hz.