Don't know if it's worth it, but one thing to do in the meantime would be to issue an error message if "-append" is used WITHOUT also specifying a file type.  The code for that is in the function parse_save_options located in the file libinterp/corefcn/load-save.cc.

The function is

string_vector
load_save_system::parse_save_options (const string_vector& argv,
                                      load_save_format& fmt, bool& append,
                                      bool& save_as_floats, bool& use_zlib)
{
#if ! defined (HAVE_ZLIB)
  octave_unused_parameter (use_zlib);
#endif

  string_vector retval;
  int argc = argv.numel ();

  bool do_double = false;
  bool do_tabs = false;

  for (int i = 0; i < argc; i++)
    {
      if (argv[i] == "-append")
        {
          append = true;
        }
      else if (argv[i] == "-ascii" || argv[i] == "-a")
        {
          fmt.set_type (MAT_ASCII);
        }
      else if (argv[i] == "-double")
        {
          do_double = true;
        }
      else if (argv[i] == "-tabs")
        {
          do_tabs = true;
        }
      else if (argv[i] == "-text" || argv[i] == "-t")
        {
          fmt.set_type (TEXT);
        }
      else if (argv[i] == "-binary" || argv[i] == "-b")
        {
          fmt.set_type (BINARY);
        }
      else if (argv[i] == "-hdf5" || argv[i] == "-h")
        {
#if defined (HAVE_HDF5)
          fmt.set_type (HDF5);
#else
          err_disabled_feature ("save", "HDF5");
#endif
        }
      else if (argv[i] == "-v7.3" || argv[i] == "-V7.3" || argv[i] == "-7.3")
        {
          error ("save: Matlab file format -v7.3 is not yet implemented");
        }
#if defined (HAVE_ZLIB)
      else if (argv[i] == "-v7" || argv[i] == "-V7" || argv[i] == "-7"
               || argv[i] == "-mat7-binary")
        {
          fmt.set_type (MAT7_BINARY);
        }
#endif
      else if (argv[i] == "-mat" || argv[i] == "-m"
               || argv[i] == "-v6" || argv[i] == "-V6" || argv[i] == "-6"
               || argv[i] == "-mat-binary")
        {
          fmt.set_type (MAT5_BINARY);
        }
      else if (argv[i] == "-v4" || argv[i] == "-V4" || argv[i] == "-4"
               || argv[i] == "-mat4-binary")
        {
          fmt.set_type (MAT_BINARY);
        }
      else if (argv[i] == "-float-binary" || argv[i] == "-f")
        {
          fmt.set_type (BINARY);
          save_as_floats = true;
        }
      else if (argv[i] == "-float-hdf5")
        {
#if defined (HAVE_HDF5)
          fmt.set_type (HDF5);
          save_as_floats = true;
#else
          err_disabled_feature ("save", "HDF5");
#endif
        }
#if defined (HAVE_ZLIB)
      else if (argv[i] == "-zip" || argv[i] == "-z")
        {
          use_zlib = true;
        }
#endif
      else if (argv[i] == "-struct")
        {
          retval.append (argv[i]);
        }
      else if (argv[i][0] == '-' && argv[i] != "-")
        {
          error ("save: Unrecognized option '%s'", argv[i].c_str ());
        }
      else
        retval.append (argv[i]);
    }

It's a reasonably straightforward change, although it would cause the code to grow quite a bit.  One would need to define a bool variable at the start, something like have_fmt, as false.  Then, in every else if branch where fmt.set_type is called you would need to set have_fmt to true.  At the end of this one could check whether "append" was true, but "have_fmt" was false and issue an error in that case.

Marking as confirmed, but changing the Summary to reflect the specific issue.

When the file type for the original save and the file type of the '-append' save match, then the operation completes succesfully.  However, if they do not match then strange errors are thrown that don't immediately point back to this root cause.

Sample code

a=1;b=2;
unlink ('ab.mat');
save -v7 ab.mat a
save -text -append ab.mat b
clear a b
load ab.mat
a
b

This code fails with

error: load: invalid element type = 8227

However, if you specify matching formats then it works correctly

a=1;b=2;
unlink ('ab.mat');
save -v7 ab.mat a
save -v7 -append ab.mat b
clear a b
load ab.mat
a
b

The Octave default save format is -text. That's why I asked what the Matlab behavior is when the preference has been changed to use -v6 as the default. Does it still append to a v7 file in v7 format, or does it now append in v6 format? Does -append also work with a v6 file without the -v6 option?

Sure, I understand the problem statement. Still hoping you can provide us with some more information so we can evaluate whether this is a bug and whether to do anything about it.

If you are instead just looking for advice or help on how to work around this difference in your code, then you should probably try the help mailing list, not the bug tracker.

It's unclear from your description whether this is because '-v7' is the default in Matlab.

What happens if you open the Preferences in Matlab and set the default "MAT-File save format" to either v6 or v7.3? Are you still able to save a file with '-v7' and then append a variable to it with '-append'?

According to Matlab's help for the save command, if the file is in v6 format, you need to use both '-append' and '-v6'. If that's true, it doesn't seem that unreasonable to require the same in Octave with '-v7'.

But it might be a useful enhancement if Octave were able to detect and use the existing file format when saving with '-append'.