Footer Links.
Do you realise how annoying it is to scroll to the bottom of the page and see a link … and then watch it shoot away from you?
And then to scroll down, and go through it again?

Unexpected behaviour.
As per above – many users are Not used to that behaviour.
Some may not scroll down far enough to trigger it … and if you haven’t included pagination (As a default), you just shot your page-views through the foot.
Some implementations are ugly – jolty, disjointed and off putting. You may find people leave because of how disorientated it makes them feel.
Slow can be a nightmare – if you are loading images and they are not compressed, not using a sharded or CDN’d location structure, you may bore users to death waiting 5+ seconds for those images.

No page views.
If you are dynamically loading content, fetching in page2+ into page1, I’m willing to bet that you aren’t logging it correctly.
If you’ve implemented it improperly, then you have skewed your stats, and may make decisions based on bad data.

Wrong/Lost link value.
People trying to link to the 5th load on pageX are going to cause problems. They think they are linking to what they just saw, where as link-followers will see the initial load only!
Further, you may find that the link value(s) point to the wrong URL.

It can be done right though.

There are use cases when I.S. is useful, even expected.
But you have to put it together right, test it and track it properly.
a) It’s an additional layer. Use standard pages and pagination as standard, then place infinite scroll over it.
b) Make it optional. Give the user the ability to enable it (or disable it).
c) Use the right script calls to fetch the content, and hit your log file with a fetch at the same time.
d) Give the user feedback – “loading page 3 of 7” is helpful.
e) Manipulate the URL with each load, and utilise hashbang/hijaxed URL structures. This permits people to link to the relevant “section” – and bots know it’s a specific page.

This way you don’t alienate users, you retain link value, you provide standard behaviours and links etc. etc. etc.
(Honestly, it doesn’t take much to figure the issues nor the solutions – it’s just a matter of making the effort before playing with a shiny new toy)

Personally I think infinite scroll is the spawn of the Devil. Tasks should have a beginning and an end. In out tests, most people gave up half-way through, no matter how enticing the content was.

I have one question which nobody can answer sensibly – WHY??