AreaGapAndOverlapCleaner does not fill gaps with grouping enabled

few things I'd check:

• Verify that all polygons surrounding the donuts/slivers belong to the same group value.
• Inspect the features that form the boundary of the hole. If they're split across multiple groups, the cleaner won't be able to resolve the gap as it did when processing the dataset globally.
• Run a simple test by assigning all features the same group value. If the holes are filled again, that would confirm the grouping is the cause rather than the tolerance setting.
• Check whether any attributes used for grouping contain unexpected nulls, whitespace, or subtle value differences that could be creating more groups than expected.

Since you're seeing the same behaviour in both FME Form 2026.2 and 2025.2.4, this feels less like a regression and more like the topology being evaluated differently once grouping is applied.

It may also be worth examining the geometry of the remaining "donuts". In some cases they're not being treated as gaps within a group at all, but as interior rings of a polygon, which can change how the transformer handles them when processing grouped features. GeometryFilter prior to areaGapAndOverlapcleaner passing each type through separately 

I think I've been able to narrow down the issue. The AreaGapAndOverlapCleaner does not seem to edit single features. To test this, I sampled 1 feature from my dataset. When the sampler followed by the AreaGapAndOverlapCleaner immediately, the slivers/donuts are not removed. When I clone the exact same feature, followed by the AreaGapAndOverlapCleaner, the slivers/donuts are filled as expected. Grouping does not influence this.

So, did I always misunderstood the way the AreaGapAndOverlapCleaner should work, or is this a bug? I think the last, because there is no other transformer which does the same for single features? (The SliverRemover does not exist any more and the SpikeRemover and DonutHoleFilter as combination is not doing exactly the same)