The slow opening of AlphaFold 3 predictions with LIG_X residue codes does not appear to be because of failure to fetch templates.  There is something very bad happening though.  I tested opening an AF3 prediction of modest size 7800 atoms with 3 ligand residues codes LIG_A, LIG_B, LIG_C and it took an incredibly slow 60 seconds to open in the ChimeraX 1.12.  If I renamed those 3 ligands to LIGAA, LIGBB, LIGCC, and restarted ChimeraX it opened in less than half a second.  In both cases the ChimeraX Log reports it failed to fetch the ligand templates.  Also if I open the LIG_A, LIG_B, LIG_C file a second time in the same ChimeraX session it opens is less than half a second.

  Greg Couch wrote the ChimeraX mmCIF reader and hopefully he can investigate and fix it.  I've attached my two example files for testing.