VeloCloud UCaaS Issues

Hello Folks,

Been having some interesting issues with our edge deployment in a retail setting. We've replaced our old routers at 80 sites with VeloCloud edges and all 80 sites now have one-way audio issues intermittently, often enough to be a headache.

1. We tried disabling the firewall (allow any any).
2. We tried either of 2 circuits only (no VCG, no load-balancing, all traffic in/out one circuit) to eliminate the chances of SDWAN related finnickiness.
3. We took PCAPs and while signalling is fine, what we do not see is RTP inbound from our UCaaS provider. Normally we'd blame them, but this is reproducibly caused by adding the VeloCloud appliance. We're waiting on the UCaaS provider to provide their logs...
4. Removing the Velo from the equation and replacing it with a standard Cisco router (same NAT, network config) solves the issue.
5. SIP ALG is disabled and SIP handshake appears to work properly.
6. Codecs are properly negotiated to G711 in both successful and failed calls.

Anyone heard of any interesting "features" from Velo? I ran a similar setup on SilverPeak without issues, and an SDWAN appliance is just a smarter router, it shouldn't really treat packets significantly differently, yet something is clearly happening.

I would guess somewhere in the SIP negotiations during bad calls, something is being modified or inserted such that RTP is not sent or sent to the wrong address, etc. Waiting on UCaaS logs to back that theory up.

Any thoughts?