First, my view:
If there's an ET, follow it. It's there for a very good reason, though the reason may not be immediately apparent.
If there's no ET, then go straight ahead. The NTOFP is analysed for terrain this way, the SID isn't. In fact, data is not available to analyse the SID track accurately in much of the world.
There is, however (and this is where it gets boring) a query over interpretation of ICAO document 8168 on this topic. It states that one should follow the 'departure route' following a malfunction on take-off. Some interpret this to mean the Standard Instrument Departure (capitalised and with that word 'Instrument'). However, there is no written clarification available, to my knowledge.
My own view is that in the immediate moments after a malfunction, climbing straight ahead is by far the safest option, and should only be deviated from where there is a safety reason not to do it. That said, a 'proper' automatic aircraft (Airbus or Fokker, for example) makes such a good job of flying the EOSID that the case is less easy to prove.
It is difficult to make a safety case for poling a Boeing around a fifty degree turn by hand at 500ft AAL immediately following a malfunction, simply because 'someone' says you should follow the SID.