Yes, it definitely could happen.
The mathematical version of gerrymandering is easily expressed in integer programming, as shown here (for people who like MIPs).
In its simplest version (without geographical considerations and vote uncertainty), the gerrymandering problem amounts to a simple partitionning problem. Let us assume you are a daredevil (there is no voter uncertainty) and want a margin of 1%. You want as many districts as possible with a percentage of voters for party A at 51 %. The number of delegates (districts) from party B has to be minimized.
In algorithmic terms, assuming building a district is like putting voters in big bags of equal size, the optimal solution is given by:
Build districts with 51 % of voters for party A and 49% of voters for party B
When you do not have enough voters for party A, build districts with voters of party B only (and the few )
This can give you for instance:
for an area with 51 % (or more) of voters for party A, 100% of delegates for party A
for an area with 25.5 % of voters for party A, 50 % of delegates for party A
and so on.
Now assume party A has a bad term, and 2 % of the voters change their minds and decide to vote for party B. If they are equally distributed between districs, you would end up with 100 % of representatives for party B. Here you have your tidal wave reversal.
In practice, the number of voters who change their minds is more likely to be decribed by a distribution function (I feel like a Rayleigh distribution would describe it well, and be used by most statisticians as a "default" distribution if there is no additional knowledge). But even if the "change of minds" differs between districts, you will have a significant shift. The worst case scenario is if the "change of minds" keeps a "gerrymandered" solution such as described above. In that case, with a minority party A, you will get twice as big a shift in representatives than in opinion change (if 2% of voters change their minds, at least 4% of representatives will be changed).
This answer concerns an "ideally gerrymandered area", politicians are usually more cautious when it comes to this sport. The margins may be larger. I also ignored the fact that districts come in finite number (4 % of representatives does not exist if you have 10 districts for instance)