This is one of those lack-of-time-and-funding free idea posts. I had this idea several years ago but have not been able to act on it with all the other priorities I have to deal with. So, I am throwing it out there in case someone who can wants to explore it. If someone has the time and resources this could be done fairly quickly to see if it might work.
Double stranded RNA triggers the RISC (RNA Induced Silencing Complex) pathway and the cell degrades single stranded RNA that matched the double stranded sequence.
Imagine inverting part of a gene so that the resulting mRNA had a reverse sequence for a stretch of nucleotides, at least 21 nucleotides, seven or eight codons, and preferably longer. This might disrupt the amino acid sequence but if chosen carefully, the right gene and position, the disruptions could be minimal with conservative amino acid changes. It also might be in a UTR without disrupting function too much (or even a "tag" extension to a UTR). (I am not sure if targeting an mRNA intron sequence would be efficient enough knockdown because splicing may occur at a high enough rate to evade degradation...?)
A homozygote for the sequence inversion could be fine. However, a heterozygote with wildtype could trigger the RISC system if the mRNAs bound together. This would knockdown expression of the corresponding gene. While it might be viable (which would help to engineer the system in the first place), it could have a very strong fitness effect.
If heterozygotes have a lower fitness than either homozygote you have underdominance, which can be utilized to transform a wild population.
Originally I was thinking along the lines of a more complex two-locus system using phiC31 integration sites for transformations; however, cassette exchange could be used to modify a single locus (e.g., Zhang, X., Koolhaas, W. H., & Schnorrer, F. (2014). A versatile two-step CRISPR-and RMCE-based strategy for efficient genome engineering in Drosophila. G3: Genes| Genomes| Genetics, 4(12), 2409-2418.).