One node ends a seeded cell on 10.4 kHz bandwidth instead of 62.5 kHz #748

What was seen

In the end-to-end runs for #746 (native MeshCore 1.16.0 and 1.17.0, scotland-strict, five seeds), every node reported sf 8 and bandwidth_hz 62500 at the end of each cell, except one: in seed 4417, AngusOutlaw1 reported bandwidth_hz 10420. It did so in every arm run with that seed, on both firmware versions.

What it is not

10420 is not a decode error. It is the SX126x's 10.4 kHz bandwidth setting (0x08 in SetModulationParams), a value the chip accepts and the bridge reads back as reported (internal/firmware/bridge.go, the BandwidthHz field of the radio stats record).

Why it looks like traffic, not the node

The reading depends on the seed and not on the arm: it appears in seed 4417 only, and in every arm with that seed. That points at something the seed decides, the traffic or the timing, retuning that one node, or a stats record caught while the modem was mid-reconfiguration, rather than at the node's configuration or the firmware build.

To find out

• Reproduce with seed 4417 and watch AngusOutlaw1's radio stats over the cell, not only at the end, to see when the bandwidth changes and whether it changes back.
• Check what the node received just before: a command or packet that alters radio parameters would explain a seed-dependent retune.
• Check whether a node on a different bandwidth for part of a cell is visible anywhere else in the results, since it would hear nothing and be heard by nothing while it is.