This project was definitely a success, albeit not a total success. This biggest weak point was, as I expected, the user. There were other weak points though, so I'll document those here in case that helps others.

These are roughly in date order

Failures seen

Intermittent power loss

Some early problems were hit which were a bit odd, but the root cause was simply that the 8xAA cell holder had some really limp springs to hold the cells in place. Even gravity could cause the cells to lose their connection, and if any knock happened it would almost certainly cause an interruption.

I never came up with a proper fix for this, but stretching the springs out periodically seemed to help. The real solution would just be a better cell holder, either with stiffer springs, or a better design. Sadly all the cell holders I could find (from ebay, Maplin, RS, etc.) all seemed to be the same poorly made item.

Logging problems

The script would logrotate at the beginning of running, then half way through it would copy the latest logs to USB drive, if present. This meant we could often end up with only a log for today (due to logrotate) and not a log for a full run (due to copying part way through the script). The easiest fix here would be to also copy off the rotated syslog files, but at no point did I hit a serious enough problem that I needed to fix that (which required updating the scripts, which was non-trivial).

User failure

The unit eventually died due to "fatal" problem. The fatal problem? Both sets of batteries died (the main power and the RTC backup battery), so things got a bit confused. Changing both sets of batteries would have fixed that.

Doh!

This is why making everything as idiot proof as possible is ideal, and remember that an idiot will always find a new way to be stupid that you didn't think of!

System inaccessibility

The top system box contained the RPi, the camera, the RTC, and extension leads running to the other box. This box was very firmly gaffer taped to a tree, to try to avoid any movement causing glitches in the time lapse output.

Future improvement

• Logging improvements, to at least beef up what feedback we had
• Find a way to fix the camera securely but still providing access when needed. Alternatively: move everything to the second box or accessible from the second box.
• Create a decent lens hood. 3D printing might be needed unless we can find something that already exists that exactly matches the field of view the camera requires.
• solar charging of the main batteries would have been a massive improvement in one respect - as then if the user couldn't get on site things should keep running if all else was well (it obviously would mean that if any problems were hit then longer would pass without detecting this, but in retrospect that is a trade off we couldn't avoid).
• if I had to use AA cells again I would custom make a better cell holder, to avoid the nonsense problem of the poorly made holder I bought.