It's fundamentally the wrong idea to rely on _fileexists to say whether or not an IO operation will succeed. The file might exist but you have insufficient permission to open/delete it, another program might remove the file between your program checking and using the file, there may be a network error (if the file is from a network share). A more robust solution is to simply catch the error and respond accordingly:
'At the top of the program
On Error Goto general_error
Dim Shared Error_happened As Long

'Main Program

End
general_error:
Print "Fatal error"; err; "at line"; _errorline
end

io_error:
Error_happened = err
Resume Next

Sub delete_files(names$)
Error_happened = 0
On Error Goto io_error
Kill names$
On Error Goto general_error
If Error_happened then
  'Ignore or inform user, depending on use case
End If
End Sub


You will have a much easier time of things if you start by assuming the input does not match, and only setting the return value to true when it does.

General note: a .a file is an archive file, which typically contains one or more .o object files. The linker is capable of pulling object files directly out of archives, otherwise you can examine them with the ar tool.

Here's the hacky method mentioned by luke:
https://qb64forum.alephc.xyz/index.php?topic=4602.msg140102#msg140102

Not quite, that requires the function name to be known as compile time. OP wants to call a function based on a string name known at runtime, which I would imagine looking up symbols in the executable in a fashion similar to loading a dynamically linked library.

No, it wouldn't have effected the size of new posts. Posts are only written to the history table when you make an edit. For a fresh post, it only exists in the live table which was already set to the larger limit.

You'll note that forum itself counts your characters. The database should be able to handle any post within that limit.

Luckily there was a separate setting to control just the inline attachments, so now it should be back to "how it was".

Le Sigh

Fixed now. The great thing about standards for escaping HTML is that there's so many to choose from...