I'll start with the Cliff Notes version of my argument. (1) CC Sabathia is no longer a big-league starter. (2) The Yankees have three southpaw relievers, but no true lefty specialist. (3) CC's decline has been entirely against righties, so he still has use as a lefty specialist reliver. (4) This isn't as weird as it sounds: other declining lefty starters have had second lives in their 30s as relievers. I'll skip the Nth discussion of how CC has been horrible for about 300 innings spanning 2013-2015, except to add that CC has zero value to the club even if he could run a 4.5ish ERA (which he can't), because the team has too many alternatives ranging from replacement-level at worst (still better than CC) to mid-rotation-level. After Michael Pineda and Nathan Eovaldi: (a) Chase Whitley and Adam Warren look like equally serviceable back-end starters; (b) another similar back-end starter is on the way in Chris Capuano, (c) Bryan Mitchell, who's looked strong in AAA, is a wild card who could be awful but could be very good; and (d) Ivan Nova should return next month, and if healthy (which he seems to be) is a solid 3rd starter or better. So even without CC, the Yankees have enough equal or better starter options now, with two more returning soon from the DL.
I love the Yankees' bullpen, but no lefty is a true specialist: Andrew Miller is the closer (or, with Betances, a terrific 8th-inninger); Justin Wilson has basically no L/R split (.594 OPS v. LHB, .604 v. RHB); Chasen Shreve has a large reverse split (.647 v. LHB, .516 v. RHB). I'm excited about Jacob Lindgren, but they drafted him #1 to be a high-leverage reliever, not a low-innings specialist.
A little-noted aspect of CC's decline is that he's as good as ever, or better, against lefties. His whole decline is from righties hitting him 200 OPS points better now:
Sabathia's OPSvRHB//OPSvLHB Career: .704//.641 2013: .804//.662 2014: .921//.570 2015: .880//.500
This isn't surprising; he's switched from power-pitcher to soft-tosser whose best pitch (by OPS against) is his changeup. Even a lively hard fastball moves much less than a changeup, so good fastballs usually have much less lefty/righty split than changeups, which slow-roll dangerously inward to opposite-handed hitters, but tail away from the same-handed. So CC's evolution makes sense: with a weak fastball, he relies more on his changeup, so he now gets hammered by righties, but he still can beat lefties, holding them to a .500s to low .600s OPS.
He'd only get better as a reliever. Most pitchers improve when they go from rotation to bullpen, for two reasons. First, they stop having to face the same hitters a second time, so they often dump their 3rd- or 4th-best pitch. Second, they no longer have to pace the arm through 90-100 pitches, so they typically gain velocity -- as we've seen with Warren, who went from the low 90s to mid 90s when moved to the bullpen, then returned to the low 90s when returned to the rotation. (I couldn't find the study proving this is almost universal, but I found one reference to a study noting that "14 of the 17 pitchers in the last 3 years with at least 60 innings as a starter and 20 innings as a reliever had a higher average fastball velocity out of the pen").
CC's old, fat, weak-kneed body makes him all the more likely to improve his pitches with the limited usage of a lefty specialist. In 2014 and 2015, CC's fastball has averaged only 89.5 MPH, but has peaked at around 92. His fastball won't ever be a real weapon again, but if as a reliever he could more consistently reach 91-92, the increased velocity separation from his changeup, which has averaged 83-84, would make both pitches more effective.
This all probably sounds weird, but it's not unprecedented for a declining lefty starter to have a second life as a lefty specialist asked to tax his arm less (with fewer innings) and to accomplish less (i.e., just get out lefties). Examples: • LHP Darren Oliver: quality starter through age 28 (106 ERA+); terrible starter, 29-33 (82 ERA+); outstanding reliever, 35-44 (149 ERA+). • LHP Greg Swindell: quality starter through age 27 (112 ERA+); mostly bad starter, 28-30 (82 ERA+); outstanding lefty reliever, ages 32-35 (142 ERA+).
I'm sure there are others, but my point isn't to do a roll call of all folks like this; my point is that there really is precedent for an early/mid-30s transition from collapsed starter to quality lefty reliever.
Do the Yankees have room to replace him? Sure, the roster spot for a new starter (Mitchell now, or Capuano or Nova soon enough) can be created by dumping Esmil Rogers. Rogers is a righty, but more precisely, he's a nothing. His 2.60 ERA is small-sample-size garbage: his 4.46 FIP shows he's the same craptastic pitcher who's far too homer-prone to succeed; you're of no use, in any role, when your HR/9 rate is 1.6 for two years. His sole value is that in a blowout or as of the 14th inning, he can eat innings, because he's versatile as a crap starter and equally crap reliever. But CC, even if moved into a mainly low-innings lefty-specialist role, can also fill that role of Sporadic Eater Of Low Leverage Innings; a recent starter moved to the pen can still toss 50-60 pitches if needed.
I'm not under any delusion this is in the works. I'm actually writing this because I haven't heard the idea floated (but I'll be happy if it has been!). It may seem weird, but given what CC has degenerated into, the much weirder option may be leaving him in the rotation for months more, when there are plenty of more solid alternatives -- and a possible useful limited role for CC in the pen.