My personnal one would be "do not overdo it", I see some people (and lets be honest, on some parts I was like them at first) that put a lot of time and efforts into:
-optimising every single event to run faster (changing conditions and actions constantly, which without measures or experience can become the opposite.)
-trying to use things they do not understand instead of either learning about it or simply not using it (the a = lerp(a,b,x*dt) is the perfect exemple of thing that is not only wrong when the calculation is done with dt variations, but also is mostly used just because it looks good rather than being needed or understood), it can demand quite a time to polish some of them pretty well
-learning by themselves how things work by doing some studies of them, whan the manual actually does this job, sure, there is a difference between reading it and proving it, but if people want to prove everything every time, it takes time for only rarely benefits.
-reinventing the wheel, seriously, why use an engine if you reinvent the wheel constantly, there are tools, and they are pretty much working in most cases, might as well learn to give them some work sometimes.
the lost time is actually time not improving the game in most cases, and the result obtained can make things worse, which means basically even more lost time.
on the same topic, backups are great to erase big f*ckups and mistakes we do, use them.