@laursen many thanks for the insights. But the issue is actually not ‘how to recover from 429 errors’, but rather ‘how to prevent 429 errors’.
The 429 error comes when the server rejects an API call because it thinks you have made too many calls already. See log below. And you can only recover from this error when the server decides to once more accept calls. (Apparently sometimes it decides rather soon to accept calls again, and sometimes it takes longer before the server will decide to accept calls again…)
2022-07-22 11:30:10.637 [TRACE] [ng.gardena.internal.GardenaSmartImpl] - >>> GET https://api.smart.gardena.dev/v1/locations, data: null
..
2022-07-22 11:30:10.867 [TRACE] [ng.gardena.internal.GardenaSmartImpl] - <<< status:429, {"message":"Limit Exceeded"}
So, as I said before, the issue is why do we get 429 errors? And how to prevent that?
The binding normally makes a couple of API calls on start-up to get the locations and initialize the web socket. Then during normal operation no further calls are made. Instead the binding relies on the server to make call-back notifications over the already established web socket. Normally the web socket should stay open so long as the token remains valid (which should be 24 hours), however in fact the server invalidates the token and drops the web socket every 2 hours. Which requires a couple of API calls again to get a new token and open a new web socket. In total therefore, the binding is making around 50 API calls per 24 hours. And we have seen over many weeks of operation that the server does not complain about that number of calls. However since the last few days, the server IS now complaining about too many API calls. Why?