Sorry to be dumb but got a couple of questions. Do the impurities we are talking about (mdp2p, mdma dimer, etc...) not crystallise? If they do, is it separate crystals and if yes, how do you separate?
a lot of people don't really understand recrystalization and why it works. The theory of recrystalization is fairly simple but there is a lot of nuance to it in practice but I'm going to simplify horribly to explain.
if you hypothetically had a mixture of 2 similar compounds one 80% and the other 20% and you dissolve them in the minimum amount of hot solvent, then one assumes that the 80% material is saturated (no more can dissolve in that solvent at that temperature), the 20% material is only 20% saturated.
The hot solution is then filtered, keeping it hot. removing all insolubles.
When you cool the filtered solution, now called the mother liquor, the 80% material will start to crystallize out pretty much immediately, the 20% material will not crystallize until it becomes saturated, (it becomes saturated as the solution cools and the solution will have to cool more before it starts to crystalize). so if the solution is cooled and stood then only a small amount of the 20% material will have crystallized out, dropped out when the mixture became saturated for the 20% material but a lot more of the 80% material will have crystallized out because that was saturated at the start. the 20% material will mostly still be in the liquid, the mother liquor. Some of the 80% material will also be in the mother liquor but most will have crystallized.
the crystals are filtered free from the mother liquor and washed with cold solvent, this removes contaminants stuck on the surface of the crystals. The resulting crystals will contain less 20% material because a lot of the 20% material is still in the mother liquor. so the crystals will be purer.
If you repeat the process enough times then there is even less 20% material at the start, so it probably will not ever crystallize, it will not ever reach saturation. so the material crystalising is practically pure. The old school metric was to recrystalize until the melting point of the crystals didn't increase, that is then pure.
A lot of people make elementary mistakes with recrystallize, for example they don't filter and wash the crystals, or worse they allow the whole mixture to evaporate, which is totally pointless because all the contaminants are still there.
recrystalization and also fractional crystalization are techniques, which in skilled hands are extremely powerful and effective. It is not without problems. for example a relatively insoluble contaminant can continue through the process and crash out early in the cooling process. Things can co-crystalize. Things can have non linear solubility curves or reversed solubility curves (more soluble in cold than hot). Things can also react with the solvent and so on...
Recrystalization is a skilled art and one that has been lost to an extent because of the prevalence of column chromatography in industry and academia. There are entire companies that exist just to do work on crystalization for the pharmaceutical industry, it is an exceptionally critical part of the process and one that if it goes wrong is an expensive nightmare to fix.
Interestingly the size of crystals has no effect on the purity, quite often smaller crystals are purer as the don't occlude mother liquor in the crystal (mother liquor stuck in defects in the crystal) smaller crystals also filter better.
Anyone with a pharma background will never say something is the same on the basis of one or two tests,
it has to match or pass on every test or it is not the same,
the basics are:
Trace analytical chromatography LC or GC often hyphenated techniques LC-MS GC-MS
Quantitative chromatography-usually HPLC
IR spectrum should be identical.
NMR if it is done, should also match.
If it is a new compound then HRMS is extremely useful to positively identify the compound.
The quantitative assay should also match. (eliminates errors due to water of crystalisation and solvates also detects invisible cutting compounds-not going to publish anything about that but they do exist and certain CMOs have been caught in the past)
The appearance should be the same, color hardness etc.
The salt morph and form should also be identical.
everybody knows how to cheat, the game is to try and spot cheating.
to get an idea for a pharma spec look up the USP monograph for any drug, and USP is quite old fashioned and limited in what is specified compared to the manufacturers control file.