[tostolaza]

Problem

Transport::jobs() sets _jobs_running = true and then calls Link::tick_resources() for every active link. A resource watchdog can send a packet, for example Resource::advertise_job re-sending an advertisement that got no part request. That packet goes through Transport::outbound(), which spins while _jobs_running is set. On single-threaded targets nothing clears the flag, so the loop never returns and the hardware watchdog resets the device.

Reproduce: nRF52840 node in transport mode serving a NomadNet page. A remote client opens a Link and requests the page over a lossy LoRa path. The log ends with:

[DBG] No part requests received, retrying resource advertisement...
[---] Resource::advertise_job: Re-sending resource advertisement packet
...
[DBG] Transport::outbound: jobs still running!

and the device resets after the 60 s watchdog.

Solution

Collect the active, non-closed links during jobs() and call tick_resources() after _jobs_running is cleared. This is the same pattern already used for announce retransmissions in outgoing.

Testing

• MASN V1.2 (nRF52840) on hardware: 31 minutes in transport mode with remote Links and resource transfers, no resets (before the fix, a reset at the first resource retry).
• Builds through microReticulum_Firmware: nRF52 (wiscore_rak4631, masn-ht-ra62) and ESP32-S3 (heltec_wifi_lora_32_V3).
• Native: not tested.

🤖 Generated with Claude Code

[attermann]

Thanks @tostolaza, nice find, and thanks for the clear repro and the hardware soak test. Collecting the links and ticking them after _jobs_running is cleared matches how outgoing is already handled, so this looks right to me.

One small thing I'll check before merging: the new tick loop now sits outside the try/catch in jobs(), so an exception from a resource watchdog would no longer be caught and logged there. If you'd like to wrap that loop in the same kind of try/catch, feel free; otherwise I may add it when I merge.