In current fasm (1.67.23) you can write things like

Does that make sense?

also, you can write

which will generate an "operand sizes do not match" error.

Tomasz:
What if we disallow the usage of size prefixes before registers, would that make fasm's code more complicated or could we even drop some code?

I mean, both examples are nonsense, the first one because register sizes are implicit, the second one because register sizes are fixed.

I think extra explicitness isn't a problem. You could use same argument for "mov eax, dword [abcd]".

i think it is for coders that are used to other asm compilers

but in fact this is not really usefull

twobytes db 0,1
db 2,3
mov eax,dword[twobytes]

this is usefull to override the size of the label, and by evidence, tomasz didn't see any problem to let the user writing explicit sizes in instructions
in my case it's really not a problem, i focus my mind on coding, not on compiler features

if you don't understand why you can write
mov eax,dword ebx
it's only because for compiler, the operands are threaded one by one, so, if you see mov, then it's a mov instruction, if you see eax, it's the eax register the destination, if you see dword, it means that the following operand size is a dword, even if it's not a dword
if you see ebx, it means that the source is ebx

there is no reason to add a test to see if there is need or not of dword like statement
it's to the coder to make this

Sometimes macros are used without restiction on operand type, but limiting to dword size. Such instances would be more difficult to code without the ability to overstate implicit sizes.