Answer
First confirm adequate system resources on the affected servers: CPU isn't maxed out, and there's sufficient available RAM and disk space. Then try stopping and restarting both of the relevant Windows services for one specific failed instance; if that doesn't help, run the mws-svr STOP then START commands again specifically for that same instance. This kind of startup issue is sometimes actually rooted in the NodeManager component rather than the HTTP server itself, and this stop/restart sequence commonly resolves it. Work through failed instances one at a time rather than all simultaneously, to keep startup overhead manageable and isolate which specific step actually fixed each one.