A perspective that is somewhere in between most of the current answers:
If you can afford to do this gradually (and with no failures in 18 months, it sounds like you probably can), blend this with other aspects of the job that are tied to salary. You've actually already got exactly that situation, and it will make the pill easier to swallow.
What do I mean by this? Simply that there is, in every job I've worked so far, an expectation (often unspoken, which is not good, but there none the less) that for both "technical" track folks and their direct management, the more senior you are, the higher the chance that you will get pulled into an emergency, whether or not you are officially on call.
So if you can get by with setting a "phased in" expectation that as people rise in seniority, they will (slowly) get more on-call duties, usually under very rare circumstances, and you are willing to accept that they do what your manager is already doing — to wit, not being "traditionally" on call, but just being willing to pick up the phone if work calls on a weekend — then the low failure rate may mean that you can, realistically, accept the risk of being down long enough for someone to sober up if they need to, or get back from the movie, or whatever they were doing. This may well simply be a matter of one of your most senior people being willing to take on what they currently perceive as "management responsibility" (after all, your manager is doing this right now, and honestly, I've never known a good manager who wasn't this way) as part of career growth.
Conversely, if what you need is a guarantee of someone being available to fix the problem, no matter when it happens… actually, you don't, and you can't get it anyway. There's always the possibility an asteroid / tsunami / societal collapse / whatever will wipe out your entire team at once anyway. Of course, normally that gets ignored under the logic "you're going to have bigger things to worry about, in that case". And that's fair, but it highlights the fact that this whole scenario is about risk mitigation. So instead, look at what the costs of an outage are, what your options are for dealing with mitigating the impacts, and what those costs look like, and find a balance.
Remember that from the employee's perspective, they don't care (and very likely don't know!) what "others like them" are making. Companies put a non-trivial amount of effort into trying to make sure this stays true, in fact. So to them, what some other team is making for doing some "similar" job that includes on-call pay means nothing whatsoever. Their job is the one you're proposing change, in a way that will, at best, be inconvenient and mildly problematic for them, and at worst could be wildly disruptive. And you're seriously thinking you should ask them to do this without any form of compensation, and expecting you won't lose 80% or more of them, in particular all the ones who are the "best" (because they will have the easiest time leaving)? If you really want to replace the entire team, there are far cleaner ways to go about it.
We had a situation a while back at my current employer where, due to various miscommunications, an SLA expectation was set of a group that wasn't prepared for it, more or less out of the blue. The only saving graces — and they were barely enough to keep half of the team members from sending out resumes that same day — were that:
- It was a very temporary arrangement (a couple of weeks)
- Management openly admitted that they fouled up (not "mistakes were made", but rather "I made a mistake")
- It was presented as a request that the team "come together to help weather this", much like one would with, for example, an "all hands on deck" crisis
- The team already had a strong (and actively encouraged) sense of camaraderie, such that folks trusted each other to be able to 'cover' if someone truly wasn't able to be on call, etc.
- The team was extensively reassured by the folks they would escalate problems to (myself and my boss) that while they might individually be on call on a rota, we would be available for escalations 24 x 7 — and they had several years of proof of this being true already.
That last part is more crucial than it might seem. If there is one thing worse than being "on call when you didn't negotiate for that up front", it is being in that situation and not having anyone to fall back on who is better at problem solving than you are, if you truly need it. I don't have that, but I'm also compensated for that aspect of my job — and it doesn't generally stress me out as much as most people seem to get.
But honestly? We have 24x7 operations support, "mostly 9-5" core dev support, and a few of the more senior devs are okay with getting pulled in ad-hoc — and that has proven to be good enough to support several thousand installations across the US. If you need more reliable availability than that… you really shouldn't be flinching at paying for it. If it wasn't negotiated up front, treat it as the cost of a mistake, make sure your hiring guidelines reflect what you actually need, and move on.