How flexible are number-of-pages submission guidelines for conferences?

None of the CS conferences I have ever submitted to had ever been lenient on page limits. There are often automatic checkers that will not allow you to upload a paper that exceeds the page limits. Conference guidelines are usually very clear and unambiguous, go and read them. Some conferences allow an appendix or a link to additional material that reviewers can read at their discretion.

Regarding splitting the papers - if you can do so in a clean way, and both parts make sense (tell a compelling narrative, have meaningful results) then sure. Otherwise, expect two rejections.

It will vary from conference to conference, of course, but expect the committee not to be too lenient unless the paper is extraordinary in some way. You can and should just ask. But 15 is certainly going to be easier for them to agree to than 20. You can expect that others are asking the same question, of course.

If the conference uses good reviewers prior to acceptance you might get suggestions for how to shorten it and if you accept the advice it would make it more likely to be accepted.

But I think a paper nearly half again as long as expected isn't a good candidate for acceptance. Unless it is extraordinary.

But the conference chair can give you the accurate answer.