Ok LN ops 101:
For an LNS to utilise a GW it needs to know the GW exists, and is ready and available:-
It is able to do this 2 ways - 1) it receives uplinks for nodes in range of that GW (it doesnt have to be the primary GW for that traffic, but the LNS will gather statistics none the less). Following on from that the LNS makes an assumption - that the GW is LoRaWAN compliant as there is no way to identify to the LNS that the GW may have restrictions - e.g. it could be supporting uplinks (Rx) only either by design or by accident (such as if only the Rx port has an antenna connected, or if the RF switch on a joint Rx/Tx ant has failed (it does happen!) or 2) the GW sends a periodic message - the status message - basically saying ‘Hi LNS I might have forwarded no traffic lately but I’m still here and able to send/receive’.
Secondly the LNS needs to know that the GW has available airtime capacity to allow it to Tx - send downlinks, join ACK’s, ADR or MAC commands etc. to assist in device/network optimisation and operation. It does that by maintaining track of the stats around any messages it has forwarded to the GW - again the LNS makes an assumption…that it is the only LNS in control of the GW, hence discussions on the forum around how you should not attempt to connect a GW to two networks/LNS infrastructure - as each LNS will have no knowledge of what the other might be trying to do with the GW if shared!
Any given Network will have its own (operator defined) rules of engagement that will determine how the network overall, and infrastrucrure/devices connected to that network will behave. The list is long, too long to debate here, but there are some obvious ones like, Frequency/Channel plans that are supported, how long a delay is acceptable in GW to LNS backhaul before a late packet is judged ‘too’ late and therefore dropped as a potential replay attack etc., what the safe operating margin is for link integrity with respect to how much headroom might be set when commanding ADR changes, etc., and in this context how long a LNS will consider reasonable when judging if a given GW has not been seen for ‘too’ long to then be considered viable and available to the network. This threshold is set by the Network Operator and is not determined by the GW user/owner, though there is a margin again as clearly the Internet is imperfect and messages might get dropped or lost along the way - be they traffic messages or status messages.
There has been much discussion historically on the Forum and elewhere as to what is “sensible”. After many years or working with LoRa/LORaWAN I would note that I have seen some extremes - very active/high capacity networks where loss or failure to be able to action downlinks I have seen deployed with much less than 30 sec status updates - even as low as 10secs - some GW even default to that depending on how target network profile is chosen - others with much longer status times - 60, 120 even 300+ seconds. Those networks will also have a threshold for how many ‘missed’ is considered acceptable before LNS starts to doubt GW availablility.
For TTN it is generally agreed that 30sec is an good compromise and an acceptable limit - and given the potential unreliabiity inherent in what can be a hobby network in places, often with backhaul not supported by reliable connections or SLA support, this would allow say 10 missed status messages before GW is tagged as offline. This would be 300sec, or >5 mins. I must confess I dont know exact value used by TTS(CE), but judging by how quickly GW’s get declared online/offline in the console I always assumed somewhere 5-15mins. Perhaps we ask a TTI Core team member for clarification if thread continues and there are further queries? On low traffic private networks where missed packets tolerated and accepted and where the application allows for that I have seen GW’s updating just once every 30 mins!
TL:DR for TTN there is a general agreement that 30 secs nominal is just right, 20sec or less considered too often and > 120sec, certainly >180sec too long… but at risk that using higher end may compromise coverage for other Community network users rather than just the folk setting/accepting higher elapsed times…
Advice from WHO?! As noted GW’s - as part of their stock firmware and depending on the chosen network profile selected where several are offered during GW setup - may have a default. Defaults need to be adjusted to reflect the needs and limits set by the LNW operator and are not universal!