Both ATLAS and GLOBE say my node has been up for only 8 hours 30 minutes but the 3-day graph of uptime indictaes no interruption in that period (i.e. 100% uptime).
When I run 'arm' that indicates Tor has been running for considerably longer than 3 days.
What is ATLAS/GLOBE trying to say when it presents apparently inconsistent data about a node?
UPDATE 2 October: My broadband supplier uses dynamic IP address allocation. For reasons I cannot yet understand, it changes the allocated IP address of the host running Tor every so often. The running Tor process recognises this change and rebuilds its descriptor. GLOBE eventually picks up that the IP address of my node has changed and reflects the new uptime for the node in its headlines. However it does not upgrade the graph of uptime. Not sure whether that is deliberate or a bug. The 'arm' monitor shows the changed IP address but does not change the uptime. That seems to be a bug too.