I don’t think this kind of variation counts as evolution.

First, in most evolutionary scenarios, there is a long, stable period in which “neutral variation” takes place, a species’ genome “spreads out” to include a lot of variations that don’t have any particular benefit at the time, but also no huge penalty. This is sort of in preparation for whatever the future may hold.

So, this kind of variation in the mustard, might have evolved over the last umpty-ump years, and is now being selected for.

Second, many species have been through the mill many times, and so the genome looks for (forgive the anthropomorphisms, they are not meant as they sound) environmental signals, and switches on and off features that correspond. Switching that selection on and off, is not evolution as such, the genome goes unchanged.

Third, there can be immediate selection. The seeds just gathered came from plants that have already survived a drought, so of course they were from plants better adapted. This is just an application of the two preceeding points.

The test would be to take these small plants, give them more water for a generation or two, and see if they switch back.

Better yet, sequence the genomes of the two sets of seeds and do a direct comparison - we do have that technology today - tho it’s not cheap.