Don't understand the solver??

I don't think there is a short answer to this one.

Solver calculates the EV assuming every other player plays according to "gto". But as the amount of possible scenarios is HUGE, PLO sims are never really "done". The longer the sim is calculated, the less it has "errors" (= spots that are not calculated long enough to give reliable results). Commercial softwares, like the one you used, is usually done very well so they shouldn't have errors in them. So the base point is that you should be able to trust the results.

Now the EV in 3bet pots is affected a lot by Fold Equity. KJT8 makes more straight draws and wraps than KQJ8. Like this:
KJT8 will make a wrap on AQ, Q9 and 97.
KQJ8 will make a wrap on T9.

Also KJT8 doesn't block QQxx which should fold almost 40% of times to BU 3bet (assuming the opener is at UTG).

That makes the difference. But usually we don't know how long the sim has run, maybe the EV changes a little if it would run longer. Or doesn't. That's why it's important with solvers not to focus on single hand, but look for overall strategies and reasonings, and hand categories.

P.S. for some reason solver really loves to 3bet those Qxxxds where xxx is middling cards. I haven't figured out exactly why, but I think it might be that solver needs to find a certain frequency and when it has used all the "good" hands it adds up those marginal ones. With those Qxxxds hands the EV is often quite the same with call or 3bet, but as solver needs to get enough hands to 3bet range (to reach the equilibrium) it adds those.

One more point to ones Kyyberi mentioned is about facing 4bets. When we 3bet ds hand that doesn't have A we are not really looking to fold versus 4bet. Hand like KQJ8ds doesn't flop that smooth as pointed above but it also suffer from proximity with AA. Hand like Q986ds has way less of that issue. Plus it blocks those broadway pockets less which are supposed to fold good amount.

Yeah good points, makes sense. Solver also likes kxxx$ds where the cards are middling. One thing that surprised me was how high $ds rundowns 3bet freq is as IP.

It's another question should we 3bet those hands in real tables or not.

So my though is preflop I should Keep it gto and then deviate postflop when possible. Because, for me anyways, deviating pre would expand the game tree far beyond my understanding of the game.

Deviating pre doesn't expand the game tree, if it's done on default level. If you play solver preflop, then you have default 3bet range on certain spot of X. If player pool is significantly different from solver assumptions, you can change your default 3bet to Y.

One clear example is that at lower stakes people don't fold to 3bets in general. So instead of 3betting with solver default ranges, we should think about adjusting that range to minimal FE for 3bets. If opponent's don't fold to 3bets, solver lowers the frequency and drops hands that has the most value from FE. Those are Axxx hands, where xxx is lower cards. The lower the cards are, the less we would like to see the flop after 3betting as domination becomes a negative issue. Also, players are often a bit sticky on the flop. My default change to solver ranges in those player pools is to remove those low card hands from default 3betting range.

While what you wrote is true, it's more true for RFI and OOP 3bet ranges than IP 3bet, where the solver uses many Axxx that rely much on fold equity. But when it comes to 3betting IP ranges, there aren't as much hands that rely much on fold equity.

But sometimes it's a bit confusing. like for example, QJT2 is a 3bet, but JT95/KQ89 is not.

Same sime at low stakes (5max), with FE in 3bet pots and no FE. MP RFI range is quite similar to UTG anyway.

BU vs MP open
Normal 3bet 6,2%
No FE 3bet 4,8%
22,5% difference in range

BU vs CO open
Normal 3bet 7,7%
No FE 3bet 6,1%
21% difference in range

So in position without FE solver drops over 1/5 of the range.

SB vs BU open, no FE, difference is 18%
SB vs MP open, no FE, difference is 20%
BB vs BU open, no FE, difference is 15%

So outside BB, it's pretty much 1/5 drop in range. With closer look it's quite the same stuff, frequency drops significantly with hand that has low cards.