To map more accurately, you’ll probably need to zoom in much further. To get a better overview of the area, use the “u” shortcut to turn off the data layer. Then use the “b” shortcut to switch back and forth between Bing and Mapbox (the two better resolution layers here) in the menu. You’ll then see the objects from a slightly different angle and under different lighting conditions.
Please do not ignore warnings, such as those about overlapping buildings, as in this changeset.

Please compare my shapes with yours.

osm.org/edit?editor=id&way=1547968056#map=18/-18.532016/32.139815

way/1546572313/history

The overlapping road was deleted by another mapper. (@ratrun)
Elements that intersect the boundaries of your changeset are downloaded by JOSM only if nodes they contain lie within the bounding box.
Looking at your edits, I get the impression that you tend to map outside the bbox. This always leads to problems. In JOSM, make sure that areas outside the downloaded region are shown hatched, you can find this option under the “View” tab.
Please don't just suggest focusing on buildings; instead, limit this project exclusively to buildings. So far, I've noticed that too many mistakes are being made.

changeset/186772920

changeset/186777420

squared all the buildings,
removed incorrect landuse polygons and barrier,
removed unnecessary tracks (short driveways)

This project primarily focuses on buildings

Please do not draw new roads over existing ones.

Just yesterday, I fixed several of these issues.

The turning lanes appear inconsistent.
Perhaps it would be better to remove the traffic island instead.

The data you appear to be trying to import is of extremely poor quality. Please note that most of the existing buildings were mapped as part of the “Eliminate Malaria” campaign. The project imagery at that time exhibited significant misalignments. As a result, many “Rapid” objects are actually duplicates with slightly different coordinates. Furthermore, both the “Rapid” data and the existing data appear to be outdated. Please remove these building outlines or verify them manually.

osm.wiki/Import/Guidelines

Translated with DeepL.com (free version)

way/448348789

its a waste water plant
most of the building footprints need to be retagged

Please check the wiki:

man_made=wastewater_plant

please use Rapid with care

It looks like you accidentally dragged the waterway

fantasy roads?
way/1364546941
way/1364321032

I don't have local knowledge either. But I would only consider it official if it were paved like the eastern part.

Most will probably end up as wrong-way drivers.

Service road to the filling station is correct. But it is not a through way to Makhado.

changeset/172313200

Are you serious?

This connection to a roundabout is impossible.

To get around the restrictions, enable expert mode (in the "View" tab) and download the data via the Overpass API. You can also use the Query Wizard to specify the type of data you want to download. Be careful when deleting nodes in incomplete data sets to avoid conflicts during upload.

mostly correct,
verified by a local mapper

Hi Ēka_nambara,
this looks fictional to me,
like most of your recent edits.
How are you able to survey all this different places?

Die Sentinel Bilder werden alle paar Tage aktualisiert

Hier ist der Link zu Maproulette:

https://mpr.lt/browse/challenges?query=Africa

Die Fehler werden in meist auf die Schnelle behoben ohne sich ausreichend mit der Situation zu beschäftigen. Dazu gibt es ein Leaderboard das die Teilnehmer noch mehr anspornt.

von mir vor kurzem kommentiertes changeset:

changeset/169824497

diese #fixmapafrica maproulette challenges sind absolut überflüssig und kontraproduktiv

osm.wiki/Sentinel-2

https://browser.dataspace.copernicus.eu

I can only identify two buildings here. The building you repaired is only about 4 m² in size. I switched back and forth between Bing and Mapbox imagery to get a better picture of the situation. Most of the errors addressed in this Maproulette challenge are typical beginner problems. When looking after these problems you will always find more issues nearby. A good pointer for areas that need a lot of rework. So please be careful, to not only obscure other problems, that are much more difficult to detect automatically.