Keep in mind that office computers are usually quite slow by themselves. Loading more and evaluating the code makes it even worse.

Well, there is no reason for the bundle size to grow. You simply spread the weight across different modules. If ModuleA and ModuleB share ModuleC. ModuleC is extracted into a separate chunk. Shared between both modules.

Also, by having fine grained feature modules you are less likely to import unused modules. This way you are potentially eliminating dead code as well.