ah! Are you still running the same setup? If so, you won’t probably care about Arduino drivers anyway.
However, I would like to encourage you to do so! Besides making a humble start with Protect code when accessing shared resources and ESP32 multicore task scheduling and its watchdog timers (WDT), these topics are still underrated in general, so I’d like to take the chance to write some more background information about that topic.
Recommendation to use realtime systems for reading values from digital sensors
Coming from your efforts you are currently putting into the climate chamber thing to get things absolutely right™, you should really consider moving away from Linux to a realtime system to improve data acquisition quality. Or do you actually run any RTOSsy thing on your Pi Zero already? If you are doing so, please forget about the background information I’d like to outline here.
When running an off-the-shelve Linux, you will inevitably run into fishy things, especially when increasing the sampling frequency, as the regular Linux Kernel will not play fair games when it comes to scheduling your measurement task away (it’s a multitasking system at all!) when it’s just about to read from the scale.
This is just bad in both high-frequency and low-frequency sampling scenarios because you will never be able to compensate for that as the task scheduling usually is pretty indeterministic from the perspective of the user-space programs.
While there are chances things can be slightly improved by adjusting the task scheduler of the Linux Kernel appropriately (e.g. by beginning to increase the task priority), it’s still not an RTOS system.
I don’t wont to scare you about these details, but they are there while maybe hidden in plain sight. I just want to be straight on that as others have been burnt by similar issues as well.
Just my two cents and thanks again for the efforts you are putting into that!