Well,
After removing some counters cycles on a node, I can state I have the MySensors GW working non-stop for 24h now with 2 nodes without any noticeable issue. Both nodes are using Encryption and Signing.
The nodes are simply reporting temperature and Humidity every minute or so. Will try now the behavior with actuator nodes (relay).
In the log I can see it isn't perfect yet. There are quite a few signing failures. The node #80 is 6 meters away from the gateway with line-of-sight, which concerns me a bit because when I place the nodes in the final locations, I'm sure the radio path will not be better than this. I'm assuming that the signing is failing due to radio issues - I might be wrong.
Dec 03 17:46:12 DEBUG TSF:MSG:READ,80-80-0,s=20,c=1,t=0,pt=7,l=5,sg=1:14.0
Dec 03 17:46:12 DEBUG SGN:BND:NONCE=326B170BBDFD3CE66C91DE485E2C465D98434C6917C4D7269DAAAAAAAAAAAAAA
Dec 03 17:46:12 DEBUG SGN:BND:HMAC=602409810E55FA10445022FEECCB08697878B9F0FD1652B9303FF9D543B240FD
Dec 03 17:46:12 DEBUG SGN:VER:OK
Dec 03 17:46:12 DEBUG GWT:TPS:TOPIC=mysensors-out/80/20/1/0/0,MSG SENT
Dec 03 17:46:13 DEBUG TSF:MSG:READ,80-80-0,s=40,c=3,t=16,pt=0,l=0,sg=1:
Dec 03 17:46:13 DEBUG SGN:SKP:MSG CMD=3,TYPE=16
Dec 03 17:46:13 DEBUG SGN:SKP:MSG CMD=3,TYPE=17
Dec 03 17:46:13 DEBUG TSF:MSG:SEND,0-0-80-80,s=255,c=3,t=17,pt=6,l=25,sg=1,ft=0,st=OK:<NONCE>
Dec 03 17:46:13 DEBUG SGN:NCE:XMT,TO=0
Dec 03 17:46:13 DEBUG TSF:MSG:READ,80-80-0,s=40,c=1,t=1,pt=7,l=5,sg=1:63.0
Dec 03 17:46:13 DEBUG SGN:BND:NONCE=854FF116C7AA8566A8176AAA1A319BE95D90D8417AA92EDB3EAAAAAAAAAAAAAA
Dec 03 17:46:13 DEBUG SGN:BND:HMAC=C54BAC9E3FE050E9170CEE17AB23836A98EA270ADFC089A4523E27610D587E88
Dec 03 17:46:13 DEBUG SGN:VER:OK
Dec 03 17:46:13 DEBUG GWT:TPS:TOPIC=mysensors-out/80/40/1/0/1,MSG SENT
Dec 03 17:46:14 DEBUG TSF:MSG:READ,80-80-0,s=21,c=3,t=16,pt=0,l=0,sg=1:
Dec 03 17:46:14 DEBUG SGN:SKP:MSG CMD=3,TYPE=16
Dec 03 17:46:14 DEBUG SGN:SKP:MSG CMD=3,TYPE=17
Dec 03 17:46:14 DEBUG TSF:MSG:SEND,0-0-80-80,s=255,c=3,t=17,pt=6,l=25,sg=1,ft=0,st=OK:<NONCE>
Dec 03 17:46:14 DEBUG SGN:NCE:XMT,TO=0
Dec 03 17:46:14 DEBUG TSF:MSG:READ,80-80-0,s=21,c=1,t=0,pt=7,l=5,sg=1:8.0
Dec 03 17:46:14 DEBUG SGN:BND:NONCE=BA1DABAF567A20CEABF5CFC35D638E7A1A3E5CF94204D96935AAAAAAAAAAAAAA
Dec 03 17:46:14 DEBUG SGN:BND:HMAC=8735CDA75D1148784B3D73601E7A4199DB44CFD039C75584EBA9ADD24897355A
Dec 03 17:46:14 DEBUG SGN:VER:OK
Dec 03 17:46:14 DEBUG GWT:TPS:TOPIC=mysensors-out/80/21/1/0/0,MSG SENT
Dec 03 17:47:15 DEBUG TSF:MSG:READ,80-80-0,s=20,c=3,t=16,pt=0,l=0,sg=1:
Dec 03 17:47:15 DEBUG SGN:SKP:MSG CMD=3,TYPE=16
Dec 03 17:47:15 DEBUG SGN:SKP:MSG CMD=3,TYPE=17
Dec 03 17:47:16 DEBUG TSF:MSG:SEND,0-0-80-80,s=255,c=3,t=17,pt=6,l=25,sg=1,ft=0,st=OK:<NONCE>
Dec 03 17:47:16 DEBUG SGN:NCE:XMT,TO=0
Dec 03 17:47:16 DEBUG TSF:MSG:READ,80-80-0,s=20,c=1,t=0,pt=7,l=5,sg=1:14.0
Dec 03 17:47:16 DEBUG SGN:BND:NONCE=EF0F98582847EC766BDBC6FB13B7920FCB68CA984C600DA0FBAAAAAAAAAAAAAA
Dec 03 17:47:16 DEBUG SGN:BND:HMAC=7AB74FA813953F8965393306FFF67FCA429DCE7F293E08357B9AAE4B0D33C5C1
Dec 03 17:47:16 DEBUG SGN:VER:OK
Dec 03 17:47:16 DEBUG GWT:TPS:TOPIC=mysensors-out/80/20/1/0/0,MSG SENT
Dec 03 17:47:17 DEBUG TSF:MSG:READ,80-80-0,s=20,c=1,t=0,pt=7,l=5,sg=1:14.0
Dec 03 17:47:17 DEBUG !SGN:BND:VER ONGOING
Dec 03 17:47:17 DEBUG !SGN:VER:FAIL
Dec 03 17:47:17 DEBUG !TSF:MSG:SIGN VERIFY FAIL
Dec 03 17:47:17 DEBUG TSF:MSG:READ,80-80-0,s=20,c=1,t=0,pt=7,l=5,sg=1:14.0
Dec 03 17:47:17 DEBUG !SGN:BND:VER ONGOING
Dec 03 17:47:17 DEBUG !SGN:VER:FAIL
Dec 03 17:47:17 DEBUG !TSF:MSG:SIGN VERIFY FAIL
Dec 03 17:47:18 DEBUG TSF:MSG:READ,80-80-0,s=20,c=1,t=0,pt=7,l=5,sg=1:14.0
Dec 03 17:47:18 DEBUG !SGN:BND:VER ONGOING
Dec 03 17:47:18 DEBUG !SGN:VER:FAIL
Dec 03 17:47:18 DEBUG !TSF:MSG:SIGN VERIFY FAIL
We can see in the log that the same node/child have both successful signing as well as failed signing.
This signature failure isn't critical for temp/Humidity readings (even if it fails now, it will be fine in the next round), but for actuator nodes (where you really want something to be turned on/off) it will be a concern.
I haven't noticed any other 433MHz interference (but only used the garage remote a couple of times during the day).
It can be that a buggy node sketch makes the RFM69 unstable and jams the frequency. One of the side effects is that there are no entries in the GW log and therefore it seems is it down. Well, something to take into consideration when coding the nodes.
My 2 cents