Having worked in various disciplines of plastics for too long now, I don't believe storage temperature and frequency of operation has much bearing on this problem. The coefficients of expansion on something that small are hardly measurable.
Now all plastics are not equal. I once worked on a container in South Africa which was more or less a copy of a French design. To achieve the same strength using the same testing methods the SA one required 33% more plastic, both using locally sourced 'equivalent grade' materials. So, if the source of the raw material changes, so could the ultimate strength, plasticity, etc. The other material problem could be due to the use of recycled material from rejects and start up waste.
Secondly, small changes in machine settings can change the product quality and are not always apparent immediately. Similarly two supposedly identical machines may need different settings to achieve the same ends. A few degrees rise in the mould temperature, or a few of tenths of a second less cooling time will make the finished product smaller. Another problem is that the finished product straight off the machine will always be bigger as the material shrinks as it cools, so any QA checking should be done at least an hour after production.
So, the current gears are more reliable than the Poole versions of old, but are we seeing failures due to the early part of the run being accepted before QA has passed them as OK? Or is there a dodgey impression within the tool, or one machine that runs hotter than others?
There are alot of variables that this problem could be attributed to, nailing it to one in particular is another matter. But for what its worth, my recent locos have not been problem free, and although don't see use very often, are kept in the house.
Dave