you're right. Here I know the Id part and the Body part will always be like this because this is from the stack dump data set (and from scanning most of it, it look like it always is formatted this way, especially after double-checking the schema...)
Now, for decoding the "double layered" html tags from the Body (actually only one layer, the rest is just normal html): I use html2text (python version) with a couple customization I did (still not working exactly well) but from what I seen and benchmark I did with `time, it _is_ faster than other tools I tried to use