It shouldn't be "an idea" at all. Profile the application, find the hotspots, understand what the performance limiter is, and fix it.

Granted, this is all Hard Work. I understand that. But it's the right thing to do.

From memory, Office 97 had that dedicated Office shortcut bar on the desktop that it inherited from Office 95, but that was more of a proto-Quick-Launch-bar than a startup accelerator. Though because the bar necessarily needed to load some Office DLLs from disk I can see how that would have given Word/Excel/etc a modest startup boost.

Back in 1997, most developers would show a 'splash screen' as their application loaded, because of course it takes time for applications to load.

It was rumoured to have such a thing but, iirc, did not (or at least it didn't depend on one to start fast). Such rumours got started during the Slashdot era when people were comparing the performance of open source office suites like StarOffice/OpenOffice to MS Office and wondering why there was such a huge gap. The rumours went away when Wine started being able to run Office well enough to be usable, and people discovered it started just as fast on Linux as on Windows. The secret was a special in-house linker but that was a trade secret until many years later, I think.

Back then there was much less understanding in the software industry of why 90's Microsoft was so successful. A lot of people couldn't work it out and - combined with their anti-trust moves against Netscape - just assumed the whole thing was built on cheating. In reality it was a combination of really buying into GUIs and their own Windows platform early (not an obviously successful move back then), combined with having some truly wizard-level systems hackers. It's hard to understand these days because clever hacking is hardly ever a competitive advantage now, outside of maybe game engines. It can even be a disadvantage, as it causes you to focus on micro-optimization whilst your competitor is shipping another useful feature.

Windows 95 was a massive hit, but it didn't have any particularly unique killer features from the end user's perspective. Apple had similar features in theory. The gap was the quality of their kernel and toolchain. Windows made the transition from being a cooperatively multi-tasked single address space system running on a driver-less "OS" (barely more than a fancy library), to being a pre-emptively multi-tasked OS with a wealth of loadable hardware drivers, and they managed that architecture shift in a way that preserved the hard work of their ecosystem's developers. Apple failed the same transition completely and Microsoft's other competitors were big iron UNIX vendors who delivered the same stability and features only through very expensive proprietary hardware.

This new story is emblematic of Microsoft's trajectory over the years. Their apps used to beat everyone on startup time by using tricks so clever everyone assumed they'd cheated, and now their hacking is so un-wizardly they actually do resort to cheating. These days the wizard level systems hackers are all at Apple. Oh how the wheel turns.