Invalid row produced when input data is differently chunked

I encountered an issue where csv-parser would randomly append an invalid row to the parsed data, depending on how the data was chunked:

// Normally the last row:
{
  STATIONS_ID: '535',
  MESS_DATUM: '20200309',
  QN_3: '-999',
  FX: '-999',
  FM: '-999',
  QN_4: '1',
  RSK: '6.4',
  RSKF: '4',
  SDK: '4.217',
  SHK_TAG: '0',
  NM: '-999',
  VPM: '7.5',
  PM: '-999',
  TMK: '5.0',
  UPM: '85.71',
  TXK: '7.5',
  TNK: '3.1',
  TGK: '1.3',
  eor: 'eor'
}
// Randomly (differently chunked) we get another row:
{ STATIONS_ID: '' }

• Operating System: macOS Catalina 10.15.3
• Node Version: v12.10.0
• NPM Version: 6.10.3
• csv-parser: Version: 2.3.2

Expected Behavior

The same result for the same data, no matter how it was chunked.

Actual Behavior

Different results for the same data.

How Do We Reproduce?

https://gist.github.com/hs-rhelm/e95623d9f09621aed96b93b269352049

Start by running the linked reproduction gist against csv-parser 2.3.2 and compare results for differently chunked input. Trace the parser's handling of the final row and chunk boundaries. Done means identical parsed output for the same data, without an extra { STATIONS_ID: '' } row.