Question: when and how do you need to explicitly close a stream?

Generally, you don't need to. To close, it depends on the implementation details of the stream you're dealing with.

Send back an error instead of ending the stream:

  var concat = require('concat-stream')
var through = require('through')
var http = require('http')
var qs = require('querystring')

var server = http.createServer(function (req, res) {
  req
  .pipe(through(counter))
  .pipe(concat({ encoding: 'string' }, onbody))
  
  function counter () {
    var size = 0 
    return through(function(buf, enc, next) {
      size += buf.length
      if (size > 20) {
        // **** 2.   ↴
        res.end('ETOOBIG\n')
        // **** 1. send back an error instead  ↴
        // next(null, null) 
      } else {
        next(null, buf)
      }
    }) 
  }
  
  function onbody (body) {
      var params = qs.parse(body)
      console.log(params)
      res.end('ok\n')
  }
})
server.listen(5000)

No we actually get an error back when the message is too big.

Question: so you just can call res.end() and not worry about next?

If we never call next(), we're never pulling more data from the stream.