I have a few thoughts:
I have seen this and in fact have an Avaya case open. The case is new and the circumstances are as follows:
Site A and B: 406 V2 3.0.44
Site A: 2 PRI cards
1-Single-Slot B Line 5 > Provider PRI NI2
1-Dual Slot A Line 1 > Set as PRI-NI2-CPE > Connected to dedicated telephony Point to Point T1
Line 2 >currently not used.
Site B: 1 PRI card
1-Single Slot B Line 5 > Set as PRI-NI2-CO>Other end of P2PT1
No IP trunking!
An external call from Site B Line 5 travels to Site A Line 1 via P2PT1
Site A incoming routing routes calls on Line 1 trunk group to dial out short code for Site A Line 5.
When Site B dials a busy external call, the call appears to drop or disconnect.
On closer inspection Monitor shows that Site A Line 5 received a disconnect - cause code - busy from the central office. Site A Line 1 receives a disconnect- cause code - release from Line 5. Site B Line 5 receives a disconnect- cause code - release from Site A Line 1.
So it appears to be broken by not passing relevant codes between Lines. In this case we see that IPO appears not to host true Tandum mode. Instead Site A "places" the call for Site B. Recall where Site A Line 5 passes the correct cause code however Site A Line 1 does not interprete the correct code and fails to generate a busy signal down the audio path to Site B.
Sounds buggy to me. This should work!
So.. Pri?
Have you looked at Layer 1, 2, 3 Monitor traces for disconnect cause codes?