It's more the hyper-optimising of plans by not letting teams finish their research node so a finisher can come in that is a concern. It leads to much larger, messier votes than is necessary.
I don't think it's fair to call reacting to the hyper-punishment of cutting off almost the entirety of a teams research output appropriately hyper-optimising.
If we can get to a situation where a small number of discrete changes are required each research turn I think it can all become a lot more accessible to people.
You generally can't reduce complexity by introducing additional rules, without effectively removing older rules. In particular the proposed change will make the next vote completely inaccessible. That is, I think I could maybe come up with a reasonable (but no where near optimal) research plan that takes locking properly into account. But research would have to be planned out 2 tiers ahead, to avoid risking either a team being locked into a low priority project while a higher priority project is opened up, or being idle while waiting for the higher priority project to become available. I estimate that I would take about 5-10 times as long as I did for my previous plan, so 10-20 hours instead of 2. Maybe Briefvoice could write a spread-sheet and also come up with a plan. But I think I would not be able to understand Briefvoice's plan, nor they mine, without putting a few additional hours into it. No one else would be able to understand either (unless they also did their own plan first).
A semi-hypothetical example: Say I already did the work to identify all priorities in the next two tiers and ordered them (I haven't actually done that yet, just picked a few top priorities). One currently open computing project (A) opens up two of the highest priorities in the next tier (B and C). Obviously I put Daystorm on A, and will put it on B afterwards, but if I also want to research C I need to prepare for Starfleet Science Academy to be available then. For that I need to estimate how long A will take (not easy with how inspiration works, I will get a probability distribution instead of a single date), and then estimate how long each of the possible projects for Starfleet Science Academy would take (across two research categories). I will also need to take into account how important these projects are, and how I weight project C being started late, compared to SFSA being idle for a year or two. And that's with just 2 teams. Since everything is related with everything else the difficulty is exponential in the number of teams + projects, so coming up with an optimal plan is already completely impossible, but even just considering the priorities in the next tier a little bit already makes everything extremely complicated.
And while later turns would be easier than the next one half a dozen teams (usually) being available still makes it more complicated than now, because you still have to look at when all the other projects that are being worked on are finished and still have to weight risk of being idle vs locked in too long.
Currently picking is still relatively easy, you just have to go down the list of currently open priorities, in order, and can completely ignore everything lower in priority when assigning any particular team. If you also have to take into account locking that means you always have to take into account how long the team will be locked in, and that means you have to always look at multiple projects, not just the next highest priority, when assigning a team. Or we have to ignore the entire future tech tree and that teams might be used for their other categories in the future and vote more or less like we did in the old system. That would waste most of the work you did in plotting out the research tree.
At the very least being locked in should not apply to generic teams, because that will usually make assigning the second specialization impossible (few projects will be possible to finish with a generic team in less than 10 years), and it would more or less locked out generic teams being used for anything that actually matters, lest we risk blocking a capable team out of researching that.
Also, 5pp is a really extreme penalty, basically equivalent to forbidding it completely. 1pp, 5rp or the team sitting out a turn would all be more reasonable (if you want it to happen very rarely), or 1rp if you want to make it something to be avoided, but not necessarily to organize your plans around avoiding it.
