When voice is digitized for VoIP, the continuous analog waveform is analyzed and broken up into chunks that become data packets. Each chunk is typically 20 milliseconds long (50 packets per second). These packets are referred to as RTP (real-time protocol).
When a person speaks, these packets are sent out in a continuous stream one after the other, one packet per 20 milliseconds.
On the receiving end, the packets are expected to arrive in order, where they are converted back into a continuous analog stream and delivered to the earpiece in the receiver's handset.
Packets traversing the Internet can be delayed randomly for various reasons, or lost completely. There is a "jitter buffer" in the receiving software that inserts a bit of delay so that the stream to the receiver is at the 20-millisecond rate even if the inbound pacing is slightly off after traversing the Internet. Likewise, if a packet is lost, the jitter buffer substitutes a similar sound based on the previous and following packets so that the person doesn't notice the dropout.
If the jitter (difference in transport delay between packets) is extreme, it is possible for a packet to be delayed for more than the 20-millisecond inter-packet delay and arrive after the next "chunk" of speech. It would be very confusing to the person listening to the voice if a missing chunk of speech is re-inserted after subsequent chunks have been played.
The Adtran device keeps track of these packets arriving out of sequence, discards them due to the confusion they would cause if played, and counts them as errors. "Out of order" in this case means "out of sequence" as opposed to "not functioning".
The cause of out-of-order RTP packets is typically improperly configured load-balancing between two or more different bandwidth links somewhere along the Internet transport path.
