If you can reproduce it in a small project then I will probably be able to solve it.
Thanks for the offer Erel but this just keeps getting weirder.
I was initially running a set up like this:
...............user on browser
............................... |
..........................HTTPS
............................... |
...............................V
AWS Application Load Balancer
...............|................................|
........HTTPS.....................HTTPS
...............|................................|
..............V..............................V
WebServer01........WebServer02....<--- both on same AWS EC2 instance, using LetsEncrypt
so I could avoid having to use LetsEncrypt (just another potential failure point) I transitioned my code and set up to:
...............user on browser
............................... |
..........................HTTPS
............................... |
...............................V
AWS Application Load Balancer
...............|................................|
........HTTP.......................HTTP....<--- change
...............|................................|
..............V..............................V
WebServer01........WebServer02....<--- both on same AWS EC2 instance, NOT using LetsEncrypt
which is apparently a quite common setup.
But now during testing on 4 devices (2 iPhones and 2 Pixels) the problem seems to have disappeared.
Is this a possible result or am I just premature in my reporting?