Hi @steirx,
have you thought about splitting your Node-RED instance into multiple ones?
I have been using this approach for several years now for my home systems, splitting different aspects of my system into separate instances after experiencing similar issues with some custom nodes (which where a lot more frequent back then).
E.g. instances for:
- the "core" controller
- this is running the essential stuff, so I'm very conservative regarding updates and custom nodes used
- a "bridge" instance
- makes other (non-essential) devices or services controllable via MQTT, like AVR, TV, snapcast, MPD
- Telegram bot
- Dashboards (several instances for NR Dashboard, uibuilder)
- playground for testing stuff
All communicating via MQTT and a local Redis server.
Some advantages:
- more failure resistant if something crashes your runtime
- update separately
- run different versions of Node-RED and/or nodes
So mostly, I don't even have to touch the "core controller" anymore when adding new parts to my system.
Apart from other benefits, it could also help to narrow your issue down to a single instance. It would require some redesign of your system, of course, but I think it's worth it in the long term.