Right now on ANSAX-L, the Anglo-Saxonists' listserv, there is an argument raging about the historicity of Beowulf (Yes, some things never change).
One part of the argument goes like this:
Witega 1: We have a poem in a manuscript from the 10th century. Read that poem. As it is. Don't try to reconstruct an earlier version.
Witega 2: But there are things in the manuscript that make no sense. You've got to emend them, and when you do, you reconstruct an earlier version of the poem.
Witega 1: What you think is a mistake may not be, but if it is, it may just be the equivalent of the typo and does not give you license to go creating ancestral versions of Beowulf.
Witega 2: But the scribes of Beowulf consistently get names of people and places bollixed up. The scribes don't have any idea who the Merovingians are, or the Heruli, or Eomer, or Heardred, but the poet did. So "read the poem we have" is impossible.
Witega 1: No it's not.
Witega 2: Yes it is.
And then things quickly devolve into the Argument Sketch
or, sometimes, the Fish-slapping Dance.
Perhaps Seinfeld has something to say that might be useful. The end of the famous "Bubble Boy" episode turns on a Trivial Pursuit card. The question reads "Who invaded Spain in the 8th Century?" The Bubble Boy answers "The Moors," but on the card is printed "The Moops," and so George holds that "The Moors" is wrong. Hilarity ensues.
I think reading Beowulf 'as we have it' is like insisting upon "The Moops." George may be technically correct within the rules of the game (that what is printed on the card has final within-game authority), but he is obviously incorrect in any reasonable, meaningful sense. Likewise, I think, no one really believes that the error of "dryhten wereda" for "dryhten Wedera" or "hea rede" for "Heardred" or the complete botch the scribe made of "Merovingian" reflects what the poet wrote or what the audience of Beowulf ever heard or read or even what a scribe had in front of him. And in fact, the scholars who say "read the poem we have" don't actually keep those obvious errors in their readings, because if you keep the manuscript readings as they are, you can't actually read the poem.
So the argument is really about how much to emend, and that's a reasonable and important argument to have. Should we be reconstructing possibly lost names of tribes when the philology is inconclusive, as in "egsode eorl" at the very beginning of the poem? The safe emendation is to "eorlas," though accepting that emendation implies that the scribe made a huge and obvious grammatical blunder in the sixth line of the poem; the other alternative, that the poet made a reference to the tribe of the Heruli (exemplar something like "eorle") requires us to import a lot of meaning into a place where it's not obvious from context. Where we draw the line between an error that is "obvious" and an over-clever re-writing of the poem is a very difficult question, and we need to keep debating it.
But using "read the poem as we have it" is really more rhetoric than anything else, and if you say "read the poem as we have it," you would probably, for consistency's sake, have to agree that the Moops invaded Spain in the 8th century.
[When everyone really knows that it was Bayonne, NJ, not Spain, that was invade by the Moops, and it happened in 1926]