FWA710 repeated LTE drops at ~22:00 — scp_crash_log_move_to_buf: dram warp in logs

Hi,

I'm experiencing recurring LTE disconnections on my FWA710 (outdoor installation, Finland, ISP: Elisa). The drops always happen around 22:00 local time — not every day, but whenever connection problems occur, they start at that hour. 22:00 is peak streaming time in Finland, so heavy network load and carrier aggregation reconfigurations by the operator are likely happening at that moment.

Device details:

• Model: FWA710 (S/N: S240Z35000712)
• Firmware: V1.60(ACGC.1)C0 — latest as of 11/17/2025
• Installation: outdoor
• Active CA configuration: B3 (primary, RSRP -77) + B1 (RSRP -82) + B20 (RSRP -74)

What the logs show:

Every disconnection event follows exactly the same sequence:

[SCP] scp_crash_log_move_to_buf: dram warp
[MEDMCU] reset timeout, still reset scp
[MEDMCU] reset timeout, still reset scp
ccci_mdinit: fail to open /sys/class/BOOT/BOOT/boot/boot_mode
LTE WAN ccmni1 link down
APN disconnected, inform backend.

The SCP (System Control Processor) crashes with a DRAM memory error, the modem MCU fails to reset it, and the LTE link drops. Sometimes the device recovers on its own after 15–30 seconds. But last night it happened repeatedly (5+ times between 22:00–23:15) and eventually entered a state where even a full power cycle only brought the connection back for ~21 seconds before the SCP crashed again.

The connection only recovered permanently the next morning after another power cycle — by which point the operator's network had presumably also recovered.

Recurring pattern (from UniFi WAN event log, last ~2 weeks):

The progression is always the same: High Latency → Packet Loss → Internet Down, starting at night.

Additionally, high latency spikes were recorded on Jun 3, Jun 1, May 29, May 28, May 27 — all without escalating to a full outage those times.

The Jun 4 night was the worst: 5+ full down/up cycles between 22:07 and midnight, then persistent down through the night. A hard power cycle at ~23:00 brought the connection back for only ~20 seconds before the SCP crashed again.

My interpretation: The SCP crash is triggered by something the LTE network sends during peak-load reconfigurations (possibly a carrier aggregation band change or cell handover during an active high-bandwidth session). The firmware does not handle this gracefully — it crashes instead of renegotiating.

Questions:

1. Is scp_crash_log_move_to_buf: dram warp a known issue in the MediaTek modem firmware used in FWA710?
2. Is there any fix planned, or a workaround (e.g., locking CA bands, disabling specific secondary carriers)?
3. Has anyone else seen this pattern, especially with 3-band CA (B3+B1+B20)?

As a temporary workaround I'm considering scheduling an automatic reboot at 21:55 via Nebula Maintenance to start each connection fresh before the peak hour. But that doesn't address the root cause.

Any input appreciated.