Math:
From the New Scientist article:
"A survey was circulated via the professional networking site LinkedIn, and 365 people responded. Many of them came from the FSS, but the survey also drew 65 responses from private and police labs...[later on:]Of the 21 police scientists who responded, ..."
So 365 total respondents - 65 non-FSS respondents = 300 FSS respondents, or ~82.2%
65 non-FSS - 21 police = 44 private respondents, or ~12.1%
21 police respondents = ~5.8%
(All percentages were rounded up; the values for police and private respondents were just over the 0.5% mark, so combined they would be very close to 17.8%, not 17.9%)
I would not consider 17.8% of respondents to be "a large proportion". However, digging deeper into the numbers does indicate that the non-FSS responders did, in general, agree that closing it would have a negative effect.
"Overall, 92.3 per cent of respondents said they thought the impact of the closure on criminal justice would be mostly negative, while 76.4 per cent said they thought it would lead to an increase in miscarriages of justice."
Since this isn't broken down by respondent type, the worst case we could argue is that it's possible that all 76.4% who though it would lead to more miscarriages were all in the FSS. However, even if we assume the worst, that the 92.3% is mostly from the FSS and the police, that indicates that a remainder of 5.3% of total respondents, or ~43.8% of private scientists, agree. In any case, at least 56.7% ((92.3-82.2)/17.8) of non-FSS respondents believed the impact would be negative.
That's some of the basic math. Now, if you want to get into the statistical validity of a voluntary-response survey administered through a social network, all bets are off.