In my experience, the danger of tubular spares rolling off the rim is largely overblown. I rode sew-ups exclusively until my favourite manufacturers (Clément and Vittoria) moved production to Asia. And I never felt the need to ride home on tenterhooks after mounting a spare.
First of all, don't forget that for many, many years, tubulars were all there was.
Why are these men wearing spare tyres (plural) like bandoliers? Because in the early years of the TdF, Tour director Henri Desgrange did not allow riders to receive any external assistance. Couldn't even toss them a sammie.
There was one famous incident when a rider broke his fork, pushed the bike to the next town with a blacksmith shoppe, borrowed the smithy's hearth and anvil to mend his fork, and rejoined the race (but the mend shortly also broke). Because the rules then required that no one apart the rider make any repairs to the bike.
Few of the roads were tarmac and they not uncommonly would have several punctures in a single stage, so winning the race literally often depended on continuing to ride on the rivet despite riding on a hastily-mounted spare.
As an unsupported enthusiast rider, there are mechanical measures you could take to reduce the risk, namely glue and air pressure. Some tyre glues by their nature offer better security after a roadside tyre change. The ones that dry soft and remain tacky are better in this regard that the ones that dry hard. The ones that dry hard also usually dry faster and have a higher ultimate bond strength, so are preferred by likes of track racers and criterium riders (because of the cut-and-thrust nature of their events). But they're a PITA to dismount and the glue has little to no leftover bonding strength.
And there's also tyre pressure. The more pressure you add to a tubular, the fatter
and shorter it gets, meaning it literally contracts circumferentially on the rim, rather like a Chinese finger trap.
Some of the old knowledge has been watered down or lost and IMHO where modern riders tend to go wrong is thinking they can inflate the spare to the same pressure they were running in the punctured tyre and have the same security. When the truth is if you were running six bar before the puncture, you should inflate the spare to seven or eight, or maybe nine or ten if you intend pressing your luck (another reason I carry a CO2 inflator

).
Spares also should be pre-glued (not everyone follows this practice, but then not everyone has roll-offs either), and a higher tyre pressure serves better to compress the glue on the tyre against the residual glue on the rim, resulting in a less marginal temporary bond.
I only ever used glues that remained tacky, and I always over-inflated any spare I mounted. And I only bothered moderating my speed after changing a puncture if the lean angle was so severe that I otherwise would have had to stop pedalling, because the lean angle was direct evidence of a high side-load on the tyre. AFAIK I never even came close to a roll-off.
Which isn't to say roll-offs don't happen, despite best, even professional, efforts.
https://www.youtube.com/watch?v=Gr89ku-K2WU
The most famous roll-off incident I can recall, Joseba Beloki, Stage 9, 2003 TdF. Not riding on a spare, naturally, but came down to a combination of unusual conditions. But there were upwards of 200 riders on the course that day and Beloki was the only one rolled a tyre.
[offtopic]
Merckx index, you might be interested to know that none other than Graeme Obree carried rollerblades when he competed in the 2013 international human-powered vehicle competition at Battle Mountain, Nevada. Well, not the whole set actually, just the wheels:
To me this looks uncannily like his preying mantis position carried to its logical extreme. This was on a shake-down run, he used a fully-enclosed plastic fairing in the race.
To make his vehicle as compact (and streamlined) as possible, The Flying Scotsman sought to reduce the leg articulation which ordinarily is needed to drive pedals in a circle. So he used a straight-push pedal system that converted linear motion into circular my means of bell cranks attached to crank arms on chainwheels. He used urethane wheels scavenged from a pair of rollerblades captured inside a parallel tracks (boxed in green) as the guides for his linear-push pedals.
The stark staring maddest bit of all, madder even than the 290-inch
(!!!) large gear resulting from having the first chainwheel drive a second, was the breathing apparatus. Fresh air intake came through about a 5cm hole at the bow of the fairing (effectively ram air, once at speed), connected to the engine's air intake (Obree's mouth) by means of a flexible hose scuba snorkel. Nope, not kidding.
[/offtopic]