First some data checking…
Here is the MB Release record: https://musicbrainz.org/release/301434dc-e59a-467d-b950-abfa9a2ad32e
This does NOT have a genre:

And here is the track record: https://musicbrainz.org/recording/79b64f15-7de1-4a6e-a66c-edc50671ac90
This has Modern Classical but NOT Electronic.
Here are the album artist *NSYNC genres: https://musicbrainz.org/artist/603ba565-3967-4be1-931e-9cb945394e86/tags
Just in case we are looking at track artist genres, here are the Phil Collins genres: https://musicbrainz.org/artist/401c3991-b76b-499d-8082-9f2df958ef78/tags
So still no Electronic.
The Release Group also has genres: https://musicbrainz.org/release-group/889be349-a960-4f2a-9dce-e554a306b3b1
And finally we find “Electronic” as a Genre.
So my guess is that Picard is flipping between the Release genre and the Release Group genre, but it is unclear whether this is an inconsistent MB response or an inconsistent Picard choice from consistent MB data. To attempt to work this out we need to check the MB data and see whether it is consistent or not…
I loaded the release into Picard and (assuming my settings were the same as @Tigggger 's) the API call is: https://musicbrainz.org/ws/2/release/301434dc-e59a-467d-b950-abfa9a2ad32e?inc=aliases%2Bannotation%2Bartist-credits%2Bartist-rels%2Bartists%2Bcollections%2Bdiscids%2Bgenres%2Bisrcs%2Blabels%2Bmedia%2Brecording-level-rels%2Brecording-rels%2Brecordings%2Brelease-group-level-rels%2Brelease-groups%2Brelease-rels%2Bseries-rels%2Burl-rels%2Bwork-level-rels%2Bwork-rels
We can load this URL into Chrome and look at the response. But there are LOTS of different genre-list tags in the response:
- Release Artist genres - *NSYNC - Neither Electronic or Modern Classic
- Release Group Artist genres - *NSYNC - Exactly the same as Release Artist genres.
- Release Group genres - Electronic / Pop
- Label genres - Jive - Neither Electronic or Modern Classic
- Label genres - Legacy - Neither Electronic or Modern Classic
- Track 2/8 genres - Modern Classical
As far as I can see when refreshing a few times, the MB response is the same - whilst this is not definitive, my guess is that it is Picard which is inconsistently choosing between Release Group genres and Track genres.
Second example
I haven’t done the same extensive analysis of the second example (Call My Name from 112 by 112) but Chillout is only a Release Group genre, whilst Contemporary R&B is both a Release Group and a Track genre).
Summary
I don’t have time to examine the code, but I have seen similar inconsistencies where e.g. multiple items are sent in different orders by MB for the same API call, resulting in a tag that has e.g. “Value1, Value2” on some calls and “Value2, Value1” on others, leading to exactly this sort of flip flop.
IMO, Picard needs always to be deterministic and deliver the same tags for the same underlying data, even if the sequence of data is different.
For genres this means:
- Always using the same genre source in the same circumstances
- Using the genre with the highest scoring genre
- Using e.g. the alphabetically first genre where multiple genres have the same score.
As I said, I haven’t had time to examine the Picard code, but this is where I would suggest the next investigative action should be.