it would seem that the current system has the unfortunate effect of de-incentivizing early posts to such questions
This is absolutely correct. My experience is that placing a bounty on a question most frequently gives you answers in the last few days - not the first several.
One way to overcome the problem, as a bounty-awarder
However, even if the bounty period is short, later answers will take advantage of earlier answers. Take, for example, my latest attempt to get quick answers to difficult problems:
Fast method to copy memory with translation - ARGB to BGR
I'm offering four 500 point bounties (2,000 rep total) to the best answers.
As you can see, the latest answer takes points from several existing answers and compares them so it's easier to determine the performance difference - something that particular user might not have been interested in doing, or perhaps even able to do, without viewing the answers already in place.
Now, I'm going to be particular in awarding my bounties so that all the major contributors are awarded as promised.
However, that's not the case with regular bounty problems. Frankly I'm not sure there's a good solution - in the same way that the fastest answer gets a bump (fgitw) for normal questions, the slowest answer (scite) often wins the bounties.
Conscientious bounty awarders spend time carefully evaluating all the answers, and many do take into account the earlier answers, however, as your example demonstrates, the later answer which builds on your answer is arguably better.
Options for bounty-hunters
My experience, though, is that your time is better spent on regular questions during the 7 day waiting period for a bounty than on the bounty itself. If you spend an hour you [i]might[/i] get the bounty. If you spread that hour across 5-6 questions, you are much more likely to get the same 50-100 points that you might get on the bounty.
I only attempt bounties that I know so well, that I don't need references to provide a comprehensive answer.
Further, contrary to the problem you describe in the OP, I try to answer early - first, there's less competition, and second if you get a uber-comprehensive answer in early, very few people will attempt to try to best it, since it already includes most of the angles they might approach, and theirs would obviously be duplicate work.
So, yes, it is a problem. No, there is no solution that can be technically implemented. Yes, there are some strategies that work - either answer late and enhance other contributions, or answer early with so much firepower no one else will touch it.