Your first revision (Examples of large legacy codebases that have been improved) was asking for a list of examples, which is considered "not-constructive" by StackExchange because the end result is a big list of items, each of which are equally valid and with no one-right-answer.
Your second revision (What is the best example of a large legacy codebases that have has been significantly improved) is better, however everyone's definition of "best" is different, so in order for a question asking for "the best X" to stay open, you have to clearly define what you consider "best" to be. Most efficient, best performing, etc. Otherwise the same thing happens: the end result is a list of answers that could all be equally valid, with no way for users to objectively vote which answer is better than the rest.
I've made an update to your question to try and get it reopened, by focusing on the actual problem you appear to be facing (figuring out proven ways to update a large legacy application to meet today's quality standards, without rewriting it), however it still needs 4 other reopen votes by the community to get reopened.
It's currently at 4 reopen votes, so I don't think it will have a problem getting the 5th vote and getting reopened now :)