AWS IoT Core - mqtt_helper connection was unexpectedly closed - Telstra/Simbase

We have an nRF9160 based product with firmware based on your AWS IoT Core example.

Our devices are connected to the Telstra CAT-M1 Network in Australia (50501) using Modem Firmware1.3.6.

When using a global SIM card from Simbase, we frequently get data sessions that appear to result in no data flow after approximately 20 minutes. This may repeat a couple of times, and then stay connected for a longer period, i.e. days.

On the console, we get:

00> [00:20:17.619,140] <inf> lte: RRC mode: Connected
00> [00:20:17.619,812] <inf> lte: RSRP: -103dBm (37)
00> [00:20:18.563,262] <err> mqtt_helper: Socket error: POLLERR
00> [00:20:18.563,262] <err> mqtt_helper: Connection was unexpectedly closed
00> [00:20:18.563,995] <inf> aws: AWS_IOT_EVT_DISCONNECTED

Using wireshark to look at the modem trace, every 4.5 minutes there appears to be a TCP packet sent from the modem to AWS.

At 1216 seconds (20 minutes), the device starts to perform a DNS lookup and there appears to be no response. I suspect no dataflow.

We only have this problem with Simbase (roaming). We have tried a Hologram SIM (roaming) on Telstra and a SIM homed on the Telstra Network - they both appear to operate fine with long session times. I have contacted Simbase, but they provide a response that they have lots of devices connected using their SIMs to Telstra with no issues.

I note another ticket on DevZone with something similar. When looking at the AT commands, they just happen to be using Simbase on the Telstra Network (50501)
devzone.nordicsemi.com/.../438479

Would you be able to indicate if the modem trace shows anything that could be useful in narrowing down this problem? Or if the problem is upstream of the Telstra RAN.

trace-2024-08-15T12-26-25.945Z.zip