What I noticed is the following statement :
seedrecover.py is a seed recovery tool which can take a seed with one or more mistakes in it, and recover the correct seed if not too many mistakes have been made.
How much is actually
"too many mistakes"? Would this mean that in some cases it might be impossible to come up with the correct word order or that the process would just take much longer?
That's actually the original wording from
the seedrecovery docs... AFAIK, that version doesn't offer "word order rearrangement", but is more suited to finding either a missing word (or two) or an incorrect word (or two)...
Obviously, the more mistakes the longer it will take on any given set of hardware... or, to put it another way, you will require more powerful hardware to be able to find more mistakes in a similar timeframe.
When you know the 12 words, just not the order... it's actually a lot easier as the total search space is a lot smaller... and you can kind of "shortcut" the process by discounting a lot of the options because the checksum will be invalid, so you don't need to bother doing any deriving of keys/addresses (which involves "costly" hash calculations etc)