I already fixed the incorrect locations, so here is an illustration of part of what was moved (annotating the current OSM tiles that are still representative of the incorrect locations). The blue arrows show where the POIs should have stayed.

Looks like rounding error we need to double check what's going with the code. Thanks for feedback

Ah that makes sense, I can see it now:

For example, on the x axis, it looks like it rounded to the closest 0.001

Looks like it is happening to everyone using this version, I can find the issue in changesets dating back to May this year:
https://osmcha.org/?filters=%7B%22editor%22%3A%5B%7B%22label%22%3A%22osmand%20maps%203.14%22%2C%22value%22%3A%22osmand%20maps%203.14%22%7D%5D%2C%22date__gte%22%3A%5B%7B%22label%22%3A%222020-04-01%22%2C%22value%22%3A%222020-04-01%22%7D%5D%7D

OSMCha shows me close to 1000 changesets using this version of OsmAnd.

Not sure if previous versions were also affected by the same issue.

Actually, it looks like it rounds the coordinates to 6 significant figures, not a specific number of decimals. See for example this history: https://www.openstreetmap.org/node/575564098/history

• x axis: 153.0071854 -> 153.0070000
• y axis: -27.4805811 -> -27.4806000

Older versions are also affected.

Looking at OSMCha, at least the following versions of OSMAnd are affected:

• 3.10
• 3.11
• 3.12
• 3.13
• 3.14

An example in version 3.13:

• node history: https://www.openstreetmap.org/node/4422263991/history
• changeset: https://www.openstreetmap.org/changeset/85781647

Thanks for releasing that quickly in 3.80.
I'm surprised the fix didn't make it into the release notes. Don't you think there should be some communication about this issue to get people to update as soon as possible?
Or maybe in the blog post, I you write one for the 3.80 release?

I'm part of the OSM ops team, should we be blocking any particular versions from accessing the API? What user-agent or other headers could we use to match the broken versions?

New IOS version with new User-Agent and fix is already out. The problem previous versions didn't have User-Agent specified, so this makes it tricky.
Since OsmAnd 3.80, 3.81,.... both problems are fixed