POSNER, Circuit Judge.

The district court certified this ERISA suit as a class action, dismissed one of the two claims in the suit, and granted the defendants' motion for summary judgment on the other one. The plaintiff appeals, raising issues of plan interpretation and also complaining about the district judge's refusal to allow him to conduct discovery to determine whether the plan's rejection of his claim was motivated by a conflict of interest. The only class member about whom there is information in the appellate record is the plaintiff.

He left MONY (Mutual of New York Insurance Company, now a subsidiary of AXA), where he had been employed in a senior position, in 1996. While employed there he had participated in two retirement plans. One was the Retirement Income Security Plan for Employees, which the parties call "RISPE"; it is a tax-qualified defined benefits pension plan; that is, it guarantees specified retirement benefits and provides favorable tax treatment both to the employer, who funds the plan, and the plan participants. The other plan was the Excess Benefit Plan for MONY Employees, which the parties call the "Excess Plan." It too is a defined benefits pension plan, but it is an unfunded one — that is, the benefits are paid directly by the employer rather than by a trust established and funded by the employer, and there are no special tax advantages. Such plans, which are intended for highly compensated employees, are referred to colloquially as "top hat" plans. Comrie v. IPSCO, Inc., 636 F.3d 839, 840 (7th Cir.2011); In re New Valley Corp., 89 F.3d 143, 148-49 (3d Cir.1996).

Both plans entitled the plaintiff to begin receiving the benefits promised by them when he turned 55, which he did in 2009. And both gave him a choice, to be made then, between taking his benefits in the form of a "straight life" annuity — a fixed monthly payment for the rest of his life — and taking them as a lump sum. The lump sum form was represented to be the actuarial equivalent of the annuity.

To determine actuarial equivalence requires two specifications. The first is an estimate of how long the recipient is likely to live (an estimate not challenged by either side in this case) and therefore for how long he would be likely to receive the monthly annuity payment if he chose the annuity rather than the lump sum. The second requirement is a discount rate to apply to the projected annuity payments. A discount rate is an interest rate used not to determine how an investment will grow but instead to calculate the present value of a future receipt. If (to take a simple example of how discounting to present value works) you expect to receive $100,000 20 years from now and you want to know what that's worth today and you think that interest rates over the next 20 years will be 6 percent, you can by using a present — value calculator discover that the present value of that expected future payment is $31,180.47. That is the amount that, invested at 6 percent interest compounded annually, will grow to $100,000 in 20 years. The lower the assumed interest rate, the more slowly the investment will grow and hence the higher the present value — the lump sum equivalent. At a 10 percent rate the present value of $100,000 in 20 years is only $14,864.36, while at 3 percent it would be $55,367.58. The dispute in this case is over the discount rate that the plan used to calculate the lump sum equivalent of the annuity — $1,888.46 a month — that the plaintiff was entitled to begin receiving when he turned 55.

When in 2009 the plaintiff became eligible to begin receiving benefits, he told MONY (as we'll refer collectively to the defendants, which include besides the insurance company the two pension plans in which the plaintiff participated and their administrators) that he wanted lump sums. So MONY cut him two checks. One was his RISPE lump sum, $325,054.28 (which happens to have been $10,000 less than his annual salary in his last year as an employee of MONY), and the other his Excess Plan lump sum, $218,726.38. The discount rate that the plan used to calculate his lump sum RISPE benefits was a blended rate called a "segment rate," 26 U.S.C. § 417(e)(3)(C), of roughly 5.24 percent. A segment rate is an interest rate calculated by the Treasury Department on the basis of investment-grade corporate bond rates. The details of the calculation are irrelevant to the appeal and the exact segment rate used by MONY in calculating the plaintiff's RISPE benefits is not in the record and is not a subject of dispute between the parties. The discount rate that MONY used to determine the plaintiff's Excess Plan lump sum was 7.5 percent.

The plaintiff contends that the discount rate required by both plans was a rate computed by the Pension Benefit Guaranty Corporation on the basis of annuity premiums charged by insurance companies. Applied to the plaintiff's lump sums under the two plans, this rate, called the "PBGC rate," would have been only 3 percent — less than half the average of the two discount rates that the plan used; and remember that the lower the discount rate, the greater the lump sum. (If the discount rate were zero, the lump sum would be simply the sum of the participant's predicted future benefits.) Oddly, we haven't been told how much greater the lump sums to which the plaintiff would be entitled (let alone the lump sums to which the other thousand or so members of the certified class would be entitled) would be if the lower discount rate were used. But as our numerical example indicated, the lump sums would undoubtedly be much greater.

When the plaintiff left MONY's employ in 1996, the RISPE plan provided that the discount rate would be the PBGC rate as of 120 days before the lump sum was due to be paid; and that rate turned out as we just said to be 3 percent. The Excess Plan did not specify a rate but as we'll explain it almost certainly was 7.5 percent, the rate the plan used.

A decade later, Congress, in the Pension Protection Act of 2006, Pub.L. 109-280, 120 Stat. 780, authorized plan sponsors to increase a plan's lump sum discount rate by amendment to the plan, and to make the increase retroactive if they wanted. See sections 302 and 1170 of the Act, 120 Stat. 920-21, 1063. Before the Act took effect, such a retroactive increase in the discount rate (and thus reduction in the size of the lump sum) would have violated ERISA's anti-cutback provision. 29 U.S.C. § 1054(g). The Pension Protection Act changed this but did (also in section 302) place a ceiling on retroactive rate increases for tax-qualified plans: the ceiling is the segment rate mentioned earlier. The ceiling is inapplicable to the Excess Plan because it is not tax-qualified.

In 2009, three years after the Pension Protection Act was passed and shortly before the plaintiff turned 55 and thus became entitled to begin receiving his retirement benefits, MONY raised the RISPE discount rate to the segment rate. The rate that MONY used to compute the plaintiff's benefits under the Excess Plan remained at 7.5 percent.

MONY could lawfully change the RISPE discount rate retroactively only if the plan authorized such an amendment. A plan is not required to do that, and it can if it wants promise not to, thereby creating a "contractual anti-cutback" rule that is enforceable like any other plan provision. Kemmerer v. ICI Americas Inc., 70 F.3d 281, 288-89 (3d Cir.1995). The Pension Protection Act provides an out only with respect to the statutory anti-cutback rule.

The RISPE plan in force when the plaintiff left MONY states that the pension rights of an employee who left on or before the effective date of a particular amendment to the plan "shall be determined solely under the terms of the Plan as in effect on the date of his or her termination of employment or retirement ... unless [the amendment is] made applicable to former Employees" (emphasis added). So the plan did allow MONY to amend it to change the discount rate retroactively. But the plan also provides "that no amendment shall ... reduce the Accrued Benefit of any Participant." The plaintiff was a plan participant and his benefit had "accrued" back in 1996, when he left the company. But "Accrued Benefit" is a defined term in the plan — defined as "the value of a Participant's Retirement Benefit expressed as a Straight-Life Annuity determined according to the terms of the Plan." "Retirement Benefit" is another defined term: it "means a benefit payable on the dates, in the forms" specified in a section of the plan that under the heading "Optional Forms" allows the participant to choose a lump sum "in lieu of the Normal Form," which is the straight-life annuity.

We interpret these provisions to mean that the Accrued Benefit — that which cannot be reduced retroactively by amendment — is the annuity, and that the lump sum, while a Retirement Benefit, is not the Accrued Benefit and therefore can be reduced retroactively. The term "Retirement Benefit" encompasses all forms of benefits payment that a participant can choose, including the lump sum option that the plaintiff chose in lieu of the annuity. See Call v. Ameritech Management Pension Plan, 475 F.3d 816, 820-21 (7th Cir. 2007). Nothing in the plan forbids retroactively amending the discount rate used to calculate the lump sum benefit if the participant chooses the lump sum in preference to the annuity.

The plaintiff cites our decision in Call as authority for interpreting "accrued benefit" (that which under the terms of the MONY plan can't be changed retroactively) to include a lump sum "retirement benefit." But the plan in Call had not defined "accrued benefit." And the issue in that case was not whether a lump sum pension benefit was excluded by the term "accrued benefit" but whether an early-retirement benefit, regardless of the form it took, was excluded by the term.

So the plaintiff's complaint about his RISPE benefit fails, but what about the lump sum he received as a participant in the Excess Plan? That plan isn't mentioned in RISPE. The Excess Plan is very short — eight pages, compared to RISPE's more than a hundred pages — and incorporates many provisions of the longer plan by reference. It does not specify a discount rate, as we mentioned. But it provides that benefits "shall be paid ... in accordance with an automatic payout provision of the Retirement Plan." In the definitions section of the Excess Plan we learn that "Retirement Plan" means RISPE. The parties agree that this is a directive to look to RISPE for guidance to what discount rate to use to calculate lump sum benefits under the Excess Plan. But RISPE doesn't specify an interest rate for computing lump sum benefits under the Excess Plan. What it says (in section 1.3(a)) is that "for purposes of determining lump sum distributions and for all other payment forms subject to [Internal Revenue] Code Section 417(e)" — that is, for tax-preferred plan payments — the "applicable interest rate shall be the interest rate prescribed by the Secretary of the Treasury under Code Section 417(e)," and that is the segment rate. But section 1.3(c) of RISPE provides that "for all other purposes under the Plan" the discount rate is "7.5 percent per year compounded annually."

So the question is whether the reference to "lump sum distributions" in section 1.3(a) includes benefits under the Excess Plan. If not, section 1.3(c) governs and the discount rate applicable to the Excess Plan is the "for all other purposes" rate of 7.5 percent. The latter is undoubtedly the correct reading because section 1.3(a) is limited to benefits from tax-preferred plans and the Excess Plan is not tax preferred.

The clincher to this interpretation is the plan administrators' consistent, unchallenged practice over many years of using the 7.5 percent figure to calculate lump sum benefits under the Excess Plan. When the consistent performance of parties to a contract accords with one of two alternative interpretations of the contract, that's strong evidence for that interpretation. This is a general principle of contract interpretation rather than a provision of ERISA, 2 E. Allan Farnsworth, Farnsworth on Contracts § 7.13, pp. 329-30 (3d ed. 2004); Restatement (Second) of Contracts, § 202, comment (g) (1981), but it is a principle that is applied in the interpretation of ERISA plans. See Gallo v. Amoco Corp., 102 F.3d 918, 920-22 (7th Cir. 1996); McDaniel v. Chevron Corp., 203 F.3d 1099, 1113-14 (9th Cir.2000); Allen v. Adage, Inc., 967 F.2d 695, 702-03 (1st Cir. 1992); Schultz v. Metropolitan Life Ins. Co., 872 F.2d 676, 679-80 (5th Cir.1989).

And it's no surprise that the discount rate in the Excess Plan should be as high as it is. "Top hat" plans provide gravy for highly compensated employees, and one expects them to be less risk averse than other employees, hence more likely to prefer taking their benefits in a lump sum, which they can invest in risky ventures with a high expected return — financial risk and return being positively correlated. If interest rates turned out to exceed the discount rate in the plan, the lump sum generated by the plan rate would confer a windfall on the participant, for remember that the lower the discount rate, the larger the lump sum, which the recipient can invest at whatever current interest rates are. MONY minimizes its exposure by fixing a high discount rate, which both reduces the size of its lump sum outlays and encourages plan participants to choose the annuity over the lump sum option, since, the smaller the lump sum relative to the annuity, the more attractive the annuity is.

That leaves for decision only the plaintiff's claim to be allowed discovery to determine whether a conflict of interest vitiates the rejection of his interpretation by the plans' benefits appeals committee. He thinks it suspicious that the committee upheld the ruling, initially made by a benefits administrator, on a ground different from the administrator's. There is nothing suspicious about such a sequence (which is common in adjudication) if the committee's ground is valid. But he also points out that the RISPE plan was having financial troubles in 2009 (unsurprisingly, given the state of the economy then), which required MONY to make additional contributions to the plan. And because the Excess Plan is not funded at all, the benefits payable under it come directly out of the company's pocket rather than out of a trust fund. At the oral argument the plaintiff's lawyer told us that MONY's liability to the class if the class action is successful would be in the neighborhood of $10 million — a large sum, though we haven't been told the size of either RISPE or the Excess Plan, and MONY's parent company, AXA, manages $450 billion in assets and has $18 billion in equity. See Axa Equitable, 10-Q Consolidated Balance Sheet, September 30, 2012, www.sec.gov/ cgi-bin/viewer?action=view&cik=727920& accession_number=0001193125-12-462659 (visited Feb. 19, 2013).

The plaintiff wants as a first step to see the minutes of the meeting at which the committee voted to deny his claims. But his lawyer made clear at oral argument that if he received them this would be followed by his deposing the committee's members.

We do not think that benefits review officers should be subjected to extensive discovery on a thinly based suspicion that their decision was tainted by a conflict of interest. There is a latent conflict of interest any time someone is asking for money from a company (from anyone, in fact), though it is muted to an extent if the party asking is an employee or former employee, since good relations with employees are a corporate asset. Marrs v. Motorola, Inc., 577 F.3d 783, 787 (7th Cir.2009). Formal adjudicators, such as judges, jurors, arbitrators, administrative law judges, and members of appellate boards of agencies, are largely insulated by immunity doctrines from interrogatories and depositions aimed at finding evidence of conflicts of interest. Informal adjudicators, such as members of a pension fund's benefits review committee, have a legitimate claim to a degree of similar protection from discovery, used so often as a form of harassment. Courts are drowning in discovery; imagine the burdens, not only on them but on employers, of discovery requests that must be complied with every time there is a colorable claim that private pension or welfare benefits were wrongly denied. Especially in a class action suit with a thousand or more class members, the burdens on the benefits review process of discovery in search of evidence of a conflict of interest could be considerable.

Moved by such concerns we held in Semien v. Life Ins. Co. of North America, 436 F.3d 805, 815 (7th Cir.2006), that discovery in a case challenging the benefits determination of plan administrators is permissible only in "exceptional" circumstances — circumstances in which the claimant can "identify a specific conflict of interest or instance of misconduct" and "make a prima facie showing that there is good cause to believe limited discovery will reveal a procedural defect." The continued validity of that holding has been questioned, however, see, e.g., Gessling v. Group Long Term Disability Plan for Employees of Sprint/United Mgmt. Co., 1:07-cv-483-DFH-DML, 2008 WL 5070434 (S.D.Ind. Nov. 26, 2008), in light of the Supreme Court's decision, subsequent to Semien, in Metropolitan Life Ins. Co. v. Glenn, 554 U.S. 105, 128 S.Ct. 2343, 171 L.Ed.2d 299 (2008).

Glenn is not about discovery, but it implies a role for discovery in judicial review of benefits determinations when a conflict of interest is alleged. Murphy v. Deloitte & Touche Group Ins. Plan, 619 F.3d 1151, 1161-64 (10th Cir.2010). How big a role is the question. We have interpreted the Supreme Court's opinion to mean that "the likelihood that the conflict of interest influenced the decision [of the plan administrator] is ... the decisive consideration" in whether to uphold a decision "that might just as well have gone the other way." Marrs v. Motorola, Inc., supra, 577 F.3d at 789 (emphasis in original). In other words, while "the correct standard of review to be applied [if the plan delegates interpretive authority to the plan administrator]... remains [after Glenn] the arbitrary and capricious standard, ... one of the factors that must be taken into account in applying that standard is any conflict of interest." Fischer v. Liberty Life Assurance Co., 576 F.3d 369, 375 (7th Cir.2009). And to determine the likelihood and gravity of a conflict of interest might require discovery to "identify a specific conflict of interest or instance of misconduct," a task of identification that in Semien we said was a prerequisite to discovery, not a goal of discovery.

These cases suggest a softening, but not a rejection, of the standard announced in Semien; and there can be no doubt that even when some discovery is necessary in a particular case to explore a conflict of interest, trial courts retain broad discretion to limit and manage discovery under Rule 26 of the civil rules.

With the case law in flux, this is not the occasion for our trying to trace out the contours of permissible discovery under ERISA. The reader may have noticed that in discussing the plaintiff's claim to be entitled to the lower discount rate, we said nothing about deferring to the benefits committee's decision; we sang no hosannas to discretion. We treated the claim as if it were a claim of breach of contract that had been rejected by a district court and was being reviewed by us de novo. We had no occasion to defer to a plan administrator's determination with which we might disagree — the only situation in which a deferential standard of judicial review bites. For we agreed with it.

The plaintiff could argue that if discovery were permitted and turned up evidence of a conflict of interest serious enough to vitiate the decision of the benefits appeals committee, he would be entitled to a further hearing. But a hearing before whom? Any committee composed of plan officials would have the same conflict of interest. The plaintiff would want the district court to conduct the hearing. In other words, he would want to convert this to a straightforward breach of contract case. And he would want us to review the district court's decision de novo. Well, that's what we've done; and we've concluded that even under that favorable (to the plaintiff) standard of review, which gives no weight to the decision of the benefits appeals committee, the committee's ruling must stand.

AFFIRMED.