Although this problem went away (for a time), when I last posted here in October, it has recurred several times since then, and I modified my scripts to use the legacy git method for pushing updates some time ago.
Up until last week, I’d been working on adding features to my app in the (Linux desktop) development environment for a couple of months, testing using Balena local mode and doing relatively few pushes via BalenaCloud over this period. At the end of last week, I switched back to trying to do a release on BalenaCloud, and I found that the legacy git method was nearly always failing, and the push method was failing 100% of the time.
I’ve spent about 5 days pulling my hair out, speculating whether I was going to need to pivot away from Balena as a delivery method. In the course of this period I experimented setting up a new fleet using the balena-js-hello-world repository, and finding that even with that both CLI push and git deployments were failing over 90% of the time.
Today, I saw a recent (14 days ago) message in the year old thread Unable to deploy to my fleets - #8 by mani13jha and I have confirmed that the workaround proposed by @mani13jha of using sysctl to disable ipv6 works for me too both on the Balena example fleet and my main application fleet.
I believe that it should be possible for me (at my location on with a static IP address on AussieBroadband’s NBN internet feed in Western Australia) to reproduce the problem at will, if anyone from Balena is interested in digging any deeper into it.
For me, it may not be a big issue, I’m OK with leaving IPv6 turned off at least while working on releases for this project.