re: "HR2P, spanning residues 1251–1286 in the HR2 domain, with low or no toxic effect in vitro" - HR2P is a synthetic peptide (small molecule made of a short sequence of amino acids) designed to interact like a puzzle piece with a specific segment of viral protein (a much larger molecule also comprised of a sequence of amino acids) that is essential for host cell recognition/fusion. The domain 1251-1286, which spans 35 amino acids, is a small but key segment of a large surface protein on healthy human respiratory cells that needs to be exposed for a virus to dock like a puzzle piece on the surface of a healthy cell for entry. HR2P resembles a key portion of a healthy cell's "viral docking site" and chemically fits and binds to viruses' domain needed for cell infection.
The HR2P-M2 is just a further synthetically modified peptide where they introduced an additional amino acid (Glu, Lys, or Arg) into the peptide sequence. Doing that slightly altered the chemical properties of the peptide to make it more chemically stable and soluble for use as a storable and deployable "medication" that can be administered. Probably more importantly the stability modification probably allows it to "hang around" longer waiting for a virus to show up before it degrades, and to be more soluble which helps it get to and be where viral targets will arrive after administration.
This approach is opposite the drug "camostat mesylate", which we heard about a couple weeks ago that blocks the ACE2 receptor, the viral docking site on the surface of human cells. So instead of gumming up the surface receptors of our healthy cells, the HR2P-M2 peptide only gums up the viruses' docking receptors, which seem like a better approach. However a combination of the two approaches could further inhibit cell infection.