SwitchBlox BB-SWB-E-2-NDAA: troubleshooting radio/Starlink link failover

An anonymized support email exchange, reproduced in chronological order with one post per email. Names, signatures, shipping details and private order arrangements have been removed. Investigation remains open; the later auto-negotiation explanation is a hypothesis, not a verified resolution.

Customer question — email 1 of 10

We’re using your SwitchBlox (BB-SWB-E-2-NDAA) in our ground station and
have run into an issue I’m hoping you can help us fix.

Setup:

  • Ground side: a laptop and our ground router both connect to the
    SwitchBlox. The ground router has a Silvus IP radio on a LAN port and a
    Starlink terminal on its WAN port.
  • Air side: the air router has a Starlink terminal on WAN, a Silvus radio
    paired with the ground radio on LAN, and a camera on LAN.
  • The laptop views the camera’s video feed over either link, with failover
    between the Silvus and Starlink paths.

The issue:
When the laptop connects directly to the ground router, failover works. If
I drop the active link, the feed recovers over the other one. When the
SwitchBlox sits between the laptop and the router, the feed never recovers
after the active link drops. The only change between the two cases is the
switch.

Our working theory is that the switch is unmanaged and the laptop’s link to
it never drops, so the laptop doesn’t get the link-state event that would
make it refresh its ARP and DHCP state. It’s also possible that MAC table
aging or spanning tree handling plays a part. We’d appreciate your input on
the following:

  1. What is the default MAC address table aging time on this model, and can
    it be changed?
  2. In unmanaged mode, does the switch forward, drop, or act on spanning
    tree BPDUs?
  3. Does the switch flush its MAC table on any event, such as a port link
    change?
  4. Is there firmware or a configuration for managed mode that would let us
    shorten MAC aging, flush the table, or propagate link state? If so, where
    can we get it, and how do we load it?
  5. Have you seen this behavior in other redundant-link or failover
    deployments, and is there a SwitchBlox model or configuration you’d
    recommend for this use case?

BotBlox response · email 2 of the original 10

2026-10-23 18:28:50 UTC · Anonymized summary of the email; quoted earlier messages omitted.

BotBlox’s initial response explained that the IP175G has a default MAC aging time of 300 seconds, with 30-second or disabled aging selectable via MDIO, but no out-of-the-box BotBlox software interface for changing it. The response stated that SwitchBlox does not implement STP itself and forwards standard STP BPDUs by default. Immediate MAC-table flushing on link loss was not established from the manufacturer’s documentation, and no per-port flush command was identified.

The standard board has no populated management microcontroller, so there is no managed-mode firmware to load. Some IC settings can be changed externally over MDIO. SwitchBlox Industrial was mentioned as supporting VLAN, but changing hardware was not yet recommended before diagnosing the fault.

The initial working hypothesis was that router or application failover behavior might be involved: the switch-facing MAC addresses would normally remain unchanged during an upstream path change. BotBlox asked whether the laptop’s physical Ethernet link actually drops during successful direct-connected failover and requested these tests:

  • Disconnect and reconnect the laptop cable after failure.
  • Ping the ground router, air router and camera separately.
  • If the camera remains reachable, restart the video application.
  • Identify whether traffic is TCP or UDP, and unicast or multicast.

These were initial diagnostic hypotheses, not a confirmed root cause.

Customer test results · email 3 of the original 10

2026-10-23 18:47:17 UTC · Anonymized summary of the email; quoted earlier messages omitted.

Disconnecting and reconnecting the laptop’s Ethernet cable does not restore connectivity. Power-cycling SwitchBlox does restore it.

Failover works without SwitchBlox, and also works with another switch (a Netgear).

During failure, the laptop can ping the ground router but cannot ping the air router or camera. Both TCP and UDP have been tried, with the same outcome. The camera stream was only an example: the failure affects communication with other devices across the radio or Starlink path too.

The customer asked BotBlox to investigate further because reliable failover is required.

BotBlox follow-up tests · email 4 of the original 10

2026-10-23 19:41:01 UTC · Anonymized summary of the email; quoted earlier messages omitted.

The Netgear comparison and recovery after power-cycling pointed toward SwitchBlox or its interaction with the router. However, power-cycling also drops and re-establishes the Ethernet link to the ground router, so that alone does not distinguish a switch reset from a router-interface reset.

BotBlox requested three further checks:

  1. Leave SwitchBlox powered and disconnect/reconnect only the cable between SwitchBlox and the ground router. Does connectivity return?
  2. During failure, can the ground router itself ping the air router or camera?
  3. If supported, disable Energy Efficient Ethernet (EEE) on the router’s LAN port while leaving auto-negotiation enabled, then repeat the test.

BotBlox also requested the router and comparison-switch models. EEE was raised as a possible compatibility factor based on prior investigations involving the IP175G, not as an established cause of this fault.

Customer follow-up results · email 5 of the original 10

2026-10-23 22:24:56 UTC · Anonymized summary of the email; quoted earlier messages omitted.

Reconnecting the cable between SwitchBlox and the ground router does not immediately restore the connection. Sometimes connectivity returns roughly ten minutes after reconnecting.

The ground router itself can ping the air router or camera during the failure.

The customer would investigate the EEE setting. The comparison switch is a NETGEAR GS305, a five-port unmanaged Gigabit switch.

Customer clarification · email 6 of the original 10

2026-10-23 22:27:12 UTC · Anonymized summary of the email; quoted earlier messages omitted.

The customer confirmed that their router does not support EEE.

BotBlox proposes comparative hardware tests · email 7 of the original 10

2026-10-26 03:47:18 UTC · Anonymized summary of the email; quoted earlier messages omitted.

BotBlox proposed a possible chipset-level compatibility issue involving link negotiation or packet forwarding. This remained a hypothesis.

SwitchBlox is a 10/100 Mbps switch, whereas the NETGEAR GS305 is Gigabit. That difference should not itself prevent failover, but testing different switch silicon could help narrow down the interaction.

Two alternatives were offered for testing in the same setup:

BotBlox offered one of each as evaluation samples, aiming both to restore working failover and investigate whether the issue is specific to the original SwitchBlox chipset. No successful result with either replacement had yet been reported.

Customer accepts testing and reports another observation · email 8 of the original 10

2026-10-26 03:50:53 UTC · Anonymized summary of the email; quoted earlier messages omitted.

The customer accepted the sample offer.

A further test found that inserting a Netgear switch between the ground router and SwitchBlox allows failover to work.

This is a useful compatibility observation, but it does not on its own establish the underlying cause. Shipping details have been omitted.

BotBlox assessment and next step · email 9 of the original 10

2026-10-26 03:58:39 UTC · Anonymized summary of the email; quoted earlier messages omitted.

BotBlox considered an unusual auto-negotiation compatibility issue the leading explanation and noted that Ethernet negotiation behavior can differ between older silicon implementations.

This was the support engineer’s assessment, not a confirmed diagnosis. No controlled reproduction or completed root-cause investigation was reported in the email.

BotBlox said a colleague would arrange the evaluation-sample order the following day.

Customer acknowledgment · email 10 of the original 10

2026-10-26 03:59:19 UTC · Anonymized summary of the email; quoted earlier messages omitted.

The customer thanked BotBlox and acknowledged the planned sample arrangement.

BotBlox sample-order update · email 11

2026-10-26 14:53:07 UTC · Anonymized summary of the email; quoted earlier messages omitted.

BotBlox confirmed that an evaluation order had been created for one GigaBlox and one SwitchBlox Rugged. Tracking details would follow once the order shipped.

This confirms order creation, not shipment, delivery or a successful failover test. The private order reference and personal details have been omitted.