I mean, one idea is to create sub domains - each running a different version of your app...but at that point, you kind of loose control and companies would love to stick to an older version (and you'd have to support an older version).
In terms of remaining on older versions, vendor typically restrict this through SLas - For example, you might only get your Tier 1 SLA response (acknowledgement of a serious issue within 1 hour, for example), if you're on the current version, or the previous version. Once you're on an older version those SLAs often don't apply, and large organisations mostly aren't ok with that.
> Once you're on an older version those SLAs often don't apply, and large organisations mostly aren't ok with that.
Won't that just mean they'd not like the rate of progress (say daily updates for bug fixes, weekly updates for minor changes, and monthly updates for major changes), and just leave?
They might just leave - but if it's a core service they need, it's a huge amount of effort to migrate users, data, implement a new system, rewrite all their process and documentation, helpdesk manuals, etc. They generally won't mind with general fixes and updates, but if you're changing existing workflows or features, or adding new capabilities in, they'll probably want control about how they roll that our to their users.
Enterprises generally don't sign up to SaaS services through a webform and a manager somewhere punching in their Amex (although it does happen sometimes). There'll be a proper contract, with lawyers, and negotiations, etc. There'll be clauses in the contract that most enterprises don't want and they'll try to negotiate away - for example, under what circumstances can the vendor (you) terminate the service? If the Enterprise has spent heaps of money to move onto your service, they'll want assurances that you can't terminate the contract because you feel like it - and if you do, that they can sue you for damages.
As you can see it's definitely a conscious decision that you'll need to make if you want to target these Enterprise customers - it's a big investment, and often (particularly in government or if your service is going to be a core platform) you'll get contracts that commit you to 5+ years of service, but the flipside is that now you've locked in 5+ years of revenue from that customer.