Why merging CSV files goes wrong
A CSV looks like the simplest format in computing and is one of the most inconsistent. There is no header in the file saying which delimiter it uses, which character encoding it was written in, or how a value containing a comma was escaped. Every program guesses, and when two programs guess differently you get a mess that only shows up after the data is already merged.
The accent problem, which is really an encoding problem
If you have ever opened an export and seen "Regi�o" where "Região" should be, that is a file written in Windows-1252 being read as UTF-8. It is the single most common CSV complaint in Portuguese and Spanish, because most ERP systems and older versions of Excel still export in the older encoding. Here the bytes are inspected rather than assumed: UTF-8 has a strict grammar, so a file that violates it is read as Windows-1252 instead. You can also force either one.
Semicolon or comma
In countries where the decimal separator is a comma, spreadsheets export CSV using semicolons, because otherwise 1,50 would split into two columns. That is why a file from a Brazilian or European system often refuses to open properly in a tool expecting commas. The delimiter is detected from the content, and you can override it when a file is unusual — a tab-separated export saved with a .csv extension, for instance.
Headers that do not line up
The real work is not concatenating rows, it is making sure the columns correspond. Pasting one export under another when the column order differs puts dates in the amount column and nothing warns you; the totals still add up. Matching on the header name removes that risk, and the flexible option additionally ignores accents, for when one export writes "Regiao" and another "Região".
Why it runs in your browser
CSV exports are how data leaves a system, which means they are usually the whole of something: every customer, every transaction, every employee. Uploading that to a web service to concatenate it is a disproportionate risk for a text operation. Here each file is decoded and parsed in memory, and the only thing that leaves is the file you download.
What the output file looks like
The merged file is written in UTF-8 with a byte order mark and semicolons as the delimiter. That combination is deliberate: the BOM is what makes Excel open a UTF-8 file with the accents intact instead of guessing, and semicolons keep decimal commas from splitting a number across two columns. If you need commas instead, choose XLSX as the output and let Excel handle the export, or open the result and save it again with the settings your destination expects.