I'm quite surprised that in six pages of this thread, no-one's raised errors in placenames in transcriptions. I have every sympathy with transcribers who make errors in personal names which seem 'obvious' to the relatives of the person concerned. Not only do we (I count myself in this, having transcribed for FreeBMD and WAP) have to cope with enumerators, priests etc. who are obviously descended from a long line of spiders, and with damage/fading to the original documents, but also personal names come in all sorts of varieties, both official (e.g. Katherine/Catherine/Kathryn) and as a result of parental choice or semi-literacy (all the Margrets, Jhons, Sharlots and such). Trying to second-guess your way through that lot will result in a transcription that is probably further still from what the enumerator actually intended to write.
But with place names, there is a finite number of places in the UK, most of them have standardised spellings and most of them can be fairly confidently associated with one or two specific counties. Yes, for relatively small places, especially those geographically distant from the enumerator, the same problems occur as with personal names, but I've seen real howlers in spelling and county attribution with large cities that any enumerator would have heard of, and any transcriber should have. I have a relative who according to an Ancestry transcriber was born in Wolverhampton, Co. Durham. On the census page it says 'Wolverhampton, Staffs' perfectly clearly.
On changing the record: I agree with those who have said Ancestry's approach leaves something to be desired. I prefer FindMyPast's/FreeBMD's approach where it's a question of a definite transcription error. The Ancestry 'alternative' field should be left for enumerator errors, or instances of married women being enumerated under their maiden name and other such oddities. Perpetuating errors that have no historical or documentary relevance is, I agree, bonkers, as someone else phrased it.