Please have a look at the documentation, e.g. here:
https://docs.lammps.org/Errors_details.html#fast-moving-atoms
No.
Atoms are “lost” when they cannot be properly communicated between sub-domains (which includes periodic boundaries). This usually means that the atoms move too far between neighbor list builds (and thus subdomain updates), but you can have the same effect if the sub-domains change too much, e.g. by large changes in the box dimensions.
As recommended in the forum guidelines, it will be much easier to provide assistance if you also provide details about your simulation, ideally the LAMMPS version, the platform you are running on, the command line, the input file (or even better a minimal version that is just sufficient to reproduce the issue at hand and contains no other functionality) and a representative log file.