The short answer is that, scientifically speaking, it’s not reasonable to do so.
The result will be invariably inconclusive and null when the sample size is too small. It compromises the conclusions.
Often, it is indicative of a huge bias/skew. It needs to take into account factors as to which subsets are agreeing to do it and how they may be different from other subsets.
So what is a reasonable sample size for a survey? Well, it partially depends on the type of survey, complexity, how random, type of analysis & how fancy the statistics are (ie: multivariate analysis, etc), and other things. But, you asked about a general rule of thumb… So in general, most agree that the minimum sample size should be no smaller than 100… If your entire population is smaller than 100 then you need EVERYONE to take the survey, not a sample.
Maximum should generally be up to 10% of the applicable population, but probably no more than 1,000 --again, there are exceptions, but generally speaking, doing more than 1,000 usually won’t add much to the accuracy and will just be a waste of time and resources for little to no added benefit.
Again, assuming everything is well organized/structured and minimizes other bias and influence. And again, in general. There are plenty of things that can make results inconclusive, null, compromised, and not statistically significant.