Developing late recruits: Limit duplicate roles on one route and cover weaknesses across routesMinor spoilers
Late recruits can fall far behind at the same level. On a single route, avoid heavy investment in duplicate roles. Across routes, use Merge Causality to add Defense, abilities, and class mastery.
Original video: “Why can't late-joining characters be developed?” | Bilibili creator Deerran Kan Youxi.
If one route already has developed core units, invest less in late recruits with overlapping roles. Across multiple routes, train repeat characters separately for Defense, skills, and class masteries, then merge them.
The stat gap at level 35
Why are late recruits often not recommended for further training? The most direct reason is their low joining stats.
I compared two level 35 versions of 
Mu:
| Development record | Level | Total stats, excluding HP |
|---|---|---|
Not yet recruited on  Theodora's route | 35 | 135 |
Trained from the start of  Leda's route | 35 | 180 |
| Difference | 0 | 45 |
Speed alone differs by 10. With the same role and level, a gap this large naturally favors investing resources in the unit you have already developed.
Class modifiers and a hypothesis about generated stats
Level-ups are affected by personal growth, class growth, mounts, and special abilities. Reclassing also changes current stats. Explaining the gap between the two versions of Mu requires more than saying class modifiers were omitted.
Using a rough calculation for these development stages, levels 5 to 20 give 15 level-ups. 
Gladiator's total growth bonus of 20% accounts for about 3 points, followed by about 7.5 points during 
Pugilist. Changing to Gladiator adds 2 across HP and Strength, and changing to Pugilist adds a net 4. Altogether that is only 16.5, still far short of 45.
Class growth and reclass modifiers = 3 + 7.5 + 2 + 4 = 16.5
This rough calculation still does not explain the full 45-point gap. The earlier table excludes HP, while the estimated reclass gains include it, so 16.5 is not directly comparable either. The exact generation rule remains a hypothesis.
My hypothesis is that unrecruited characters generate stats differently from personally trained units, possibly based on their current class.
Prioritize core units on one route, then merge across routes
If you are following one route to the ending and already have a core team, a late recruit with the same role does not need a large catch-up investment. Develop your existing damage dealers, tanks, and support units first.
Multiple routes give late recruits another purpose. Part 3 has Merge Causality, which takes the higher value for each stat from different versions of the same person. If your main Mu has 13 Defense and another has 18, use 18. If one has sword S and the other gauntlet S, the merged version retains both. Learned abilities, class masteries, and spells can also be combined.
Train duplicate characters to cover different weaknesses
Take 
Sirocco: he joins at the start of 
Leda's route, where I need a mage, so I train magic. On 
Theodora's route, he requires Renown 5, so I train sword classes instead. On routes where he requires even higher Renown, I immediately use armored classes to improve Defense growth.
Not every version needs to carry a battle alone. Early recruits serve as core units; later copies supply abilities, mastery, or specific stats. Combine those strengths when merging.