Everything has a tolerance, nothing is ever perfect. With some expensive GPS equipment you can get that tolerance down to a couple of centimeters, but most people aren’t using that. Often, aerial imagery (with all the problems you’ve mentioned) is more accurate than the GPS traces I get, unless I have a lot of data to average together.

In the US, unless you have high-quality public domain property data to work with, I generally recommend aligning off of the USGS 3D Elevation Program data. While the USGS doesn’t officially guarantee the horizontal accuracy of this data, it’s definitely close enough for our work. Using elevation data directly for this also eliminates error caused by different aerial imagery sources using different elevation data to create the flat imagery.

Close enough is close enough.

Replace Geometry also works for node -> way. One of the nodes of the new way will be replaced with the original node, and the tags copied. Not a perfect solution for maintaining histories when changing object type, but it does help some.

I’ve found that scars from pipeline installation can be visible in USGS 3DEP data. Pipelines with a cleared cutline will be most visible, but less maintained pipelines are also often visible. I’ve even found long-distance buried telephone lines (according to old topo maps at least), but didn’t map those because I couldn’t confirm if they were still active.

There are some surviving boot scrapers near me, mostly on Colonial Georgian and Federal-style buildings. I could only remember the location of one, but I’m sure I’ll find more if I finish surveying for historical markers.

The tagging of boot_scraper=left/right/both makes sense, but it’s probably important to specify that it’s from the perspective facing the door.

The TIGER import in the US in 2007 used tags like “name=State Route 101” for unnamed roads with a route number. In OSM, that data should be stored in the ref tag instead, as “ref=NH 101”. No road is actually named “State Route 101”, so that tag should be removed when the number has been copied to the ref.

Satellite imagery can be quite busy, which makes it much easier to not see an area that you haven’t mapped, or an area that did not end up the way you thought it should. It might not apply to everyone, but it’s something that’s helped me.

A town full of address numbers is usually the more useful of the two, but if data collection is required, it will take a significant amount of time and effort. 6 is more about not trying to map house numbers and buildings and sidewalks and… all at the same time, especially over a larger area. Trying to map many topics isn’t necessarily bad, nor is trying to improve a large area. Trying to do both at the same time, I’ve found, leads to inaccuracy.